diff --git a/.DS_Store b/.DS_Store new file mode 100644 index 0000000..8902c46 Binary files /dev/null and b/.DS_Store differ diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..a85d550 --- /dev/null +++ b/.gitignore @@ -0,0 +1,16 @@ + +public/.DS_Store +.DS_Store +public/.DS_Store +.DS_Store +.DS_Store +content/.DS_Store +public/.DS_Store +public/images/.DS_Store +static/.DS_Store +*.log +public/.DS_Store +public/images/.DS_Store +static/.DS_Store +public/.DS_Store +static/.DS_Store diff --git a/README.md b/README.md old mode 100644 new mode 100755 index 28ec04f..fab7d55 --- a/README.md +++ b/README.md @@ -1,4 +1,4 @@ -# CNI Engineering Lab Website +# CNIEL Website The following website uses the Hugo Framework: (). @@ -26,7 +26,7 @@ There are four main folders that are used. - All page-specfic css styles are in the their respective layout pages. - /_default folder - index.html is where you can edit the front page - - /partials include code blocks that can be inserted into html with {{partial "partialName" .}} + - /partials includes code blocks that can be inserted into html with {{partial "partialName" .}} - /_markup/render-image.html is used to ensure that all images in the content markup files are rendered correctly in html - /singles - Specifies the layouts for special one-off pages that need special layouts @@ -43,10 +43,10 @@ Adding pages is as simple as creating a new file and filling in the information. - 'hugo new research/researchName.md' - Fill in frontmatter - Body of text and images go in the md body - +- Pages are sorted in lists by manually changed publish/edit date (this can be used to sort members and research pages) ## Miscellaneous Notes -- Filling in the 'externalLink' field in research will specify it as an external research page +- Filling in an the 'externalLink' field in research will specify it as an external research page - This page will not appear in the navbar - You only need to fill in title and shortDescription for the list page display - Any links on the members page will be automatically converted into link-buttons diff --git a/archetypes/default.md b/archetypes/default.md old mode 100644 new mode 100755 diff --git a/archetypes/members.md b/archetypes/members.md old mode 100644 new mode 100755 diff --git a/archetypes/research.md b/archetypes/research.md old mode 100644 new mode 100755 diff --git a/config.toml b/config.toml old mode 100644 new mode 100755 index a38410e..e9cc698 --- a/config.toml +++ b/config.toml @@ -1,6 +1,6 @@ -baseURL = "https://annapoliswu.github.io/CNI_Lab/" +baseURL = "https://www.eecis.udel.edu/~ajbrock/" languageCode = "en-us" -title = "CNI Engineering Lab" +title = "CNIEL - University of Delaware" sectionPagesMenu = "main" CanonifyURLs = true diff --git a/content/.DS_Store b/content/.DS_Store new file mode 100644 index 0000000..1983d8a Binary files /dev/null and b/content/.DS_Store differ diff --git a/content/members/.DS_Store b/content/members/.DS_Store new file mode 100644 index 0000000..5008ddf Binary files /dev/null and b/content/members/.DS_Store differ diff --git a/content/members/AustinBrockmeier.md b/content/members/AustinBrockmeier.md old mode 100644 new mode 100755 index 2b2b9e2..9441527 --- a/content/members/AustinBrockmeier.md +++ b/content/members/AustinBrockmeier.md @@ -1,6 +1,6 @@ --- -name: "Austin J. Brockmeier" -title: "Some Title" +name: "Austin J. Brockmeier, Ph.D." +title: "Assistant Professor" imageLink: "/images/members/AustinBrockmeier.jpg" currentMember: true @@ -8,10 +8,27 @@ date: 2018-04-28T17:07:03-04:00 draft: false --- -I am an assistant professor in the Department of Electrical and Computer Engineering (ECE) and the Department of Computer and Information Sciences (CIS) within the College of Engineering at the University of Delaware (UD), and a resident faculty in UD's Data Science Institute. +[Curriculum Vitae](/other/ajbrockmeier_CV_12_2021.pdf) +[Email](mailto:ajbrock@udel.edu) + +*What*: **Data Science, Machine Learning, and Signal Processing** +These involve the underlying mathematical analysis, design of statistical models, and software implementation of data and signal processing (filtering and neural networks) as well as optimization routines. + +* Finding discrepancies in data sets using [divergences](/research/discrepancies/) +* Selecting representations of data using [informativeness measures](/research/informativeness/) + +*Why*: To extract actionable information from complex data, especially to **understand and interface with the brain** -My research interests include data science, machine learning, signal processing, and the underlying mathematics and design of statistical models, optimizations, and algorithms. I have experience interfacing with complex data from the domains of biomedical engineering, neuroscience, and natural language processing. +* Identify meaningful features (in space or time) of the evoked neural response using [metric learning](/research/neural_decoding/) +* Find patterns of waveforms in EEGs using [neural signal processing](/research/eegs/) +* Extract information from text and use machine learning to prioritize large collections of articles using [natural language processing and text mining](/research/pico/) [Google Scholar](https://scholar.google.com/citations?hl=en&user=g_QoCQQAAAAJ&view_op=list_works&sortby=pubdate) -[Curriculum Vitae](https://www.ece.udel.edu/wp-content/uploads/2021/03/ajbrockmeier-cv-03-2021.pdf) -[Email](mailto:ajbrock@udel.edu) + +*Where*: **University of Delaware** +* Primary appointment: [Department of Electrical and Computer Engineering (ECE)](https://www.ece.udel.edu) +* Joint appointment: [Department of Computer and Information Sciences (CIS)](https://www.cis.udel.edu/) +* Both are within the [College of Engineering](https://engr.udel.edu/) +* Resident faculty in UD's [Data Science Institute](https://dsi.udel.edu/) + + diff --git a/content/members/BilalRiaz.md b/content/members/BilalRiaz.md old mode 100644 new mode 100755 index 4680008..b6cb07e --- a/content/members/BilalRiaz.md +++ b/content/members/BilalRiaz.md @@ -1,11 +1,11 @@ --- -name: "Bilal Riaz" -title: "M.S." +name: "Bilal Riaz, M.S." +title: "Ph.D. student" imageLink: currentMember: true -date: 2021-04-28T17:00:22-04:00 +date: 2019-08-28T17:00:22-04:00 draft: false --- -Unsupervised learning as optimizations with positive semidefinite matrices. +###### Unsupervised learning as optimizations with positive semidefinite matrices. {#bilal} diff --git a/content/members/CarlosCardenas.md b/content/members/CarlosCardenas.md old mode 100644 new mode 100755 index d31521b..a0796c9 --- a/content/members/CarlosCardenas.md +++ b/content/members/CarlosCardenas.md @@ -1,11 +1,14 @@ --- -name: "Carlos Mendoza Cárdenas" -title: "M.S." -imageLink: +name: "Carlos H. Mendoza Cárdenas, M.S." +title: "Ph.D. student" +imageLink: "/images/members/Carlos.jpeg" currentMember: true -date: 2021-04-28T13:12:57-04:00 +date: 2019-01-01T13:12:57-04:00 draft: false --- +**Research focus**: finding patterns in neural time series through convolutional sparse analysis. -Works with spotting brain wave patterns using neural networks. +**Goal**: to discover physiologically meaningful waveforms in multi-day continuous epileptic electrocorticographic (ECoG) recordings that can be used to build interpretable features for seizure prediction. + +**Methods**: interpretable machine learning, clustering and sparse coding for time series, supervised learning for neural data diff --git a/content/members/CesarOlivares.md b/content/members/CesarOlivares.md old mode 100644 new mode 100755 index 2057607..435456c --- a/content/members/CesarOlivares.md +++ b/content/members/CesarOlivares.md @@ -1,11 +1,12 @@ --- -name: "C. Cesar Claros Olivares" -title: "M.S." -imageLink: +name: "C. Cesar Claros Olivares, M.S." +title: "alumnus and admitted Ph.D. student" +imageLink: "/images/members/Cesar.jpg" currentMember: true -date: 2021-04-28T17:00:02-04:00 +date: 2019-05-01T17:00:02-04:00 draft: false --- -Vigilant machine learning. +###### Error-aware and vigilant machine learning. {#cesar} +3D convolutional neural networks for brain age prediction from structural measures. Optimal transport for color and style transfer. diff --git a/content/members/EdwinSalcedo.md b/content/members/EdwinSalcedo.md old mode 100644 new mode 100755 index a7a9c9e..2060d52 --- a/content/members/EdwinSalcedo.md +++ b/content/members/EdwinSalcedo.md @@ -1,10 +1,10 @@ --- -name: "Edwin Salcedo" -title: "M.B.A" -imageLink: +name: "Edwin Salcedo, M.Sc., M.B.A." +title: "Visiting Scholar (Summer 2019)" +imageLink: "/images/members/Edwin.png" currentMember: false -date: 2021-04-28T17:36:38-04:00 +date: 2019-07-01T17:36:38-04:00 draft: false --- diff --git a/content/members/EvanCurtin.md b/content/members/EvanCurtin.md new file mode 100644 index 0000000..01cdc48 --- /dev/null +++ b/content/members/EvanCurtin.md @@ -0,0 +1,11 @@ +--- +name: "Evan Curtin" +title: "Undergraduate (Summer Scholar 2021)" +imageLink: "/images/members/EvanCurtin.png" +currentMember: true + +date: 2021-06-01T21:00:00-04:00 +draft: false +--- + +###### Non-Negative Matrix Factorization as Dictionary Learning for Audio Separation diff --git a/content/members/HassanBaker.md b/content/members/HassanBaker.md old mode 100644 new mode 100755 index 491da2b..5b949af --- a/content/members/HassanBaker.md +++ b/content/members/HassanBaker.md @@ -1,11 +1,11 @@ --- -name: "Hassan Baker" -title: "M.S." -imageLink: +name: "Hassan Baker, M.S." +title: "Ph.D. student" +imageLink: "/images/members/Hassan.png" currentMember: true -date: 2021-04-28T17:00:38-04:00 +date: 2020-02-01T17:00:38-04:00 draft: false --- -Neural decoding and signal processing. +###### Graph signal processing for functional and structural neuroimaging. {#hassan} diff --git a/content/members/HauPhan.md b/content/members/HauPhan.md old mode 100644 new mode 100755 index 7a70456..ca8f938 --- a/content/members/HauPhan.md +++ b/content/members/HauPhan.md @@ -1,11 +1,11 @@ --- -name: "Hau Phan" -title: "M.S." -imageLink: +name: "Hau Phan" +title: "M.S. and Ph.D. student" +imageLink: "/images/members/Hau.gif" currentMember: true -date: 2021-04-28T17:00:52-04:00 +date: 2021-01-01T17:00:52-04:00 draft: false --- -Reinforcement learning estimation and detection. +###### Reinforcement learning estimation and detection. {#hau} diff --git a/content/members/JustinLabombard.md b/content/members/JustinLabombard.md new file mode 100644 index 0000000..e4f1d35 --- /dev/null +++ b/content/members/JustinLabombard.md @@ -0,0 +1,11 @@ +--- +name: "Justin Labombard" +title: "Undergraduate (Summer Scholar 2021)" +imageLink: "/images/members/Justin.png" +currentMember: true + +date: 2021-06-15T17:00:22-04:00 +draft: false +--- + +###### Dictionary Learning on Epileptic ECoG Waveforms diff --git a/content/members/NicolasLopez.md b/content/members/NicolasLopez.md new file mode 100644 index 0000000..def5f07 --- /dev/null +++ b/content/members/NicolasLopez.md @@ -0,0 +1,11 @@ +--- +name: "Andres Nicolas Lopez, MSc." +title: "Visiting Scholar (Summer 2021)" +imageLink: "/images/members/Nicolas.jpg" +currentMember: false + +date: 2021-06-20T17:36:38-04:00 +draft: false +--- + +Statistician working on the error modeling for synergistic machine learning. [LinkedIn](https://www.linkedin.com/in/anlopezl/) \ No newline at end of file diff --git a/content/members/YukselKarahan.md b/content/members/YukselKarahan.md old mode 100644 new mode 100755 index 2949b80..98b828a --- a/content/members/YukselKarahan.md +++ b/content/members/YukselKarahan.md @@ -1,11 +1,13 @@ --- -name: "Yuksel Karahan" -title: "M.S." +name: "Yuksel Karahan, M.S." +title: "Ph.D. student" imageLink: currentMember: true -date: 2021-04-28T16:59:19-04:00 +date: 2019-08-28T16:59:19-04:00 draft: false --- -Data science techniques for semisupervised domain transfer. +**Research interests**: Data science techniques for semi-supervised domain transfer. Using machine learning tools to understand complex data that includes analyzing and optimizing the algorithms and their implications on data. + +**Current project**: Currently working on Wasserstein-based divergence measures to understand discrepancy between distributions. \ No newline at end of file diff --git a/content/research/.neural_decoding.md.swp b/content/research/.neural_decoding.md.swp new file mode 100644 index 0000000..aa3fe2a Binary files /dev/null and b/content/research/.neural_decoding.md.swp differ diff --git a/content/research/discrepancies.md b/content/research/discrepancies.md new file mode 100644 index 0000000..058c7c1 --- /dev/null +++ b/content/research/discrepancies.md @@ -0,0 +1,28 @@ +--- +title: "Interpretable maximal discrepancies metrics for analyzing and improving generative models" +shortTitle: "Interpretable Divergences" +shortDescription: "Project funded by the Office of Naval Research" + +authors: "Office of Naval Research, Grant # N00014-21-1-2300, Principal Investigator: Austin J. Brockmeier, 4/2021–4/2024." + +imageLink: "/images/research/discrepancies.png" +externalLink: +journalLink: +pdfLink: +codeLink: + +date: 2021-04-14T00:59:28-04:00 +draft: false +--- + +## Overview +Divergence measures quantify the dissimilarity, including the distance, between distributions and are fundamental to hypothesis testing, information theory, and the estimation and criticism of statistical models. Recently, there has been renewed interest in divergences in the context of generative adversarial neural networks (GANs). While a multitude of divergences exist, they vary in their characteristics. Importantly, not all divergences are equally interpretable: a divergence between samples is considered interpretable if it directly answers the question “Which instances best exhibit the discrepancy between the samples?” + +![Where do the distributions differ.](/images/research/discrepancies2.png "Example of discrepancies. ") + +#### Outcomes +* **Y. Karahan**, B. Riaz, and **A. J. Brockmeier** "Kernel landmarks: An empirical statistical approach to detect covariate shift", *Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021)* [Paper](https://openreview.net/forum?id=Wu5hMMQ76OE), [Poster](/other/karahan_poster_distshift2021), [5 min Presentation](https://recorder-v3.slideslive.com/?share=56618&s=79a6864d-0cec-41ad-a152-09cceb1c6bf4). +* **A. J. Brockmeier**, **C. C. Claros Olivares**, M. S. Emigh, and L. G. Sanchez Giraldo "Identifying the instances associated with distribution shifts using the max-sliced Bures divergence", *Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021)* [Paper](https://openreview.net/forum?id=y_s0M6OtyH_), [Poster](/other/brockmeier_poster_distshift2021), [5 min Presentation](https://recorder-v3.slideslive.com/?share=56676&s=b55c603d-1e31-4ab2-9cd7-440e5d05d64d). + +#### Acknowledgements +These research efforts are sponsored by the Department of the Navy, Office of Naval Research under ONR award number N00014-21-1-2300. Gratefully acknowledge the support of Dr. Tory Cobb and ONR321. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Office of Naval Research. \ No newline at end of file diff --git a/content/research/eegs.md b/content/research/eegs.md old mode 100644 new mode 100755 index c8caf11..abcca8f --- a/content/research/eegs.md +++ b/content/research/eegs.md @@ -10,13 +10,13 @@ issue: 1 pages: "43-54" year: 2016 -imageLink: "/images/research/eegs1.png" +imageLink: "/images/research/eegs2.png" externalLink: journalLink: "http://dx.doi.org/10.1109/TBME.2015.2499241" -pdfLink: "http://cnel.ufl.edu/files/1447193268.pdf" +pdfLink: "/papers/brockmeier2016_learning_recurrent_waveforms.pdf" codeLink: "/code/eegs.zip" -date: 2021-04-28T22:42:02-04:00 +date: 2016-04-28T22:42:02-04:00 draft: false --- @@ -24,4 +24,8 @@ draft: false When experts analyze EEGs they look for landmarks in the traces corresponding to established patterns such as oscillatory and phasic events of particular frequency or morphology. Long records motivate automated analysis techniques. Automation techniques often require design choices such as wavelet family or number of bandpass filters. To overcome this, we explore a modeling approach that automatically learns recurrent temporal waveforms within EEG traces. The estimation is based on a multiple-input, single-output linear model with sparsely excited inputs. -![Figure 1](/images/research/eegs1.png "Figure 1") +![We assume that an EEG signal can be described by a convolutional sparse coding model. The EEG trace is approximated as an additive mixture of component signals each described by a convolution of a sparse source with a waveform. Although this is a linear model (multiple input and single output), to separate the components requires a nonlinear analysis. Additionally, when the waveforms are unknown, this blind source separation problem is even more challenging.](/images/research/eegs1.png "Convolutional sparse coding model for EEG") + +We apply the approach to various data sets to better understand the specificity and consistency of the estimated waveforms. In particular, we cluster the waveforms found on different electrodes/channels and subjects. + +![Each cluster of waveforms is described by its centroid, the waveform nearest the centroid, the spectrum of the centroid, and the spatial distribution of the originating electrodes for the cluster waveforms.](/images/research/eegs2.png "Recurrent waveforms") diff --git a/content/research/graphs.md b/content/research/graphs.md new file mode 100644 index 0000000..b40e65a --- /dev/null +++ b/content/research/graphs.md @@ -0,0 +1,29 @@ +--- +title: "Advancing machine learning for neuroimaging through topology-aware signal +processing" +shortTitle: "Graphs and Signal Processing for Neuroimaging" +shortDescription: "Project funding from a University of Delaware Research Foundation-Strategic Initiative grant" + +authors: "University of Delaware Research Foundation–Strategic Initiative, Principal Investigator: Austin J. Brockmeier, Mentor: Gonzalo Arce, 11/2019–10/2021." + +imageLink: "/images/research/brainGraph1.png" +externalLink: +journalLink: +pdfLink: +codeLink: + +date: 2020-02-01T00:59:28-04:00 +draft: false +--- +The proposed work plan is to develop machine learning techniques to work directly with graph signal processing techniques in the context of neuroimaging. The goal is to leverage information in the form of the topology of the signal sensors or measurement locations to refine the neural signal representations in order to improve the statistical power of tests for distinguishing differences between conditions or stimuli. The project’s scope includes the formulation, mathematical and statistical analysis, and initial validation of the proposed methodology. + +![First 5 harmonics (Fourier transform) of the gray matter graph.](/images/research/brainGraph2.png "Example of harmonics of gray matter. ") + +#### Outcomes +* NER'21: **Hassan Baker**, A. J. Brockmeier "Local and Sparse Linear Causal Models for fMRI Resting-State Signals" [Full-text (IEEE)](https://ieeexplore.ieee.org/document/9441242/) | [Poster PDF](/other/baker_poster_NER2021.pdf) +*Machine Learning for Health (ML4H): **Hassan Baker**, A. J. Brockmeier "Exploring latent networks in resting-state fMRI using voxel-to-voxel causal modeling feature selection" [Paper](https://arxiv.org/pdf/2111.07488.pdf), [Poster](/other/baker_poster_ML4H2021.pdf) + + +#### Acknowledgements +Research was carried out with the support of the University of Delaware Research Foundation. Hassan Baker acknowledges support from the Gore Engineering +Graduate Fellowship in Spring 2021. This research was supported in part through the use of Data Science Institute (DSI) computational resources at the University of Delaware, specifically the DARWIN system. The DARWIN computing project at the University of Delaware is supported by the National Science Foundation under Grant No. OAC-1919839. The authors thank the support from the University of Delaware IT Research Computing Group. \ No newline at end of file diff --git a/content/research/informativeness.md b/content/research/informativeness.md old mode 100644 new mode 100755 index f98ccae..79542e0 --- a/content/research/informativeness.md +++ b/content/research/informativeness.md @@ -16,7 +16,7 @@ journalLink: "http://www.jmlr.org/papers/v18/16-296.html" pdfLink: "http://www.jmlr.org/papers/volume18/16-296/16-296.pdf" codeLink: "/code/informativeness.zip" -date: 2021-03-17T21:25:29-04:00 +date: 2017-03-17T21:25:29-04:00 draft: false --- diff --git a/content/research/neural_decoding.md b/content/research/neural_decoding.md old mode 100644 new mode 100755 index 34b8aa3..99516ac --- a/content/research/neural_decoding.md +++ b/content/research/neural_decoding.md @@ -1,7 +1,7 @@ --- title: "Neural Decoding with Kernel-based Metric Learning" -shortTitle: "Neural Decoding" -shortDescription: "We pose the problem of optimizing multi-neuron metrics and other metrics for a particular neural decoding task using centered alignment, a kernel-based dependence measure." +shortTitle: "Metric Learning for Neural Decoding" +shortDescription: "Machine learning (optimizing feature weightings and projections using kernel-based dependence) for enhancing neural data analysis, applied to a somatosensory neural decoding task." authors: "Austin J. Brockmeier, John S. Choi, Evan G. Kriminger, Joseph T. Francis, and Jose C. Principe" journal: "Neural Computation" @@ -10,31 +10,25 @@ issue: 6 pages: "1080-1107" year: 2014 -imageLink: "/images/research/neuralDecoding1.png" +imageLink: "/images/research/metricLearning_thumb.png" externalLink: journalLink: "http://dx.doi.org/10.1162/NECO_a_00591" -pdfLink: "http://cnel.ufl.edu/files/1389293595.pdf" +pdfLink: "/papers/brockmeier2014_kernel_metric_learning.pdf" codeLink: "/code/neural_decoding.zip" -date: 2021-04-28T23:19:06-04:00 +date: 2014-04-28T23:19:06-04:00 draft: false --- ## Overview +Given a sample of data points, we often assume the points reside in some space in which we can measure distances between pairs of points. From these measurements we can understand which points are close to each other. Nearby points are often assumed to share characteristics. This assumption is the foundation of nearest-neighbor classification and regression as well as clustering. -When studying the nervous system, the choice of metric for the neural responses is a pivotal assumption. A well-suited distance metric enables neuroscientists to gauge the similarity of neural responses to various stimuli and assess the variability of responses to a repeated stimulus. In particular, neural spike train metrics have been used to quantify the information content carried by the timing of action potentials. While a number of metrics for individual neurons exist, a method to optimally combine single-neuron metrics into multi-neuron, or population-based, metrics is lacking. We pose the problem of optimizing multi-neuron metrics and other metrics for a particular neural decoding task using centered alignment, a kernel-based dependence measure. +The function that measures the distance between pairs of points in a particular space is called a distance metric. -## Metric learning for neural encoding models +![In the original metric space, the white gold circle is closer to three black circles (two of which are equidistant from it). In a new metric space, changes in the vertical axes contribute more to the distance. Now, two gold circles are closer than the black circles.](/images/research/metricLearning.png "Example of metric learning. ") -Replicating the experiments of "Analyzing Neural Responses to Natural Signals: Maximally Informative Dimensions" by Sharpee, Rust, and Bialek, 2004, but replacing the maximally informative direction algorithm with metric-learning. +When studying the function of the nervous system, the choice of metric for the neural responses is a pivotal assumption. A well-suited distance metric enables neuroscientists to gauge the similarity of neural responses to various stimuli and assess the variability of responses to a repeated stimulus. In particular, neural spike train metrics have been used to quantify the information content carried by the timing of action potentials. While a number of metrics for individual neurons exist, a method to optimally combine single-neuron metrics into multi-neuron, or population-based, metrics is lacking. For time-locked neural responses, we pose the problem of the supervised optimization of multi-neuron metrics and other metrics (including those for local field potentials (LFPs), which are the electric potentials measured within brain tissue). The goal is to tune the metric to better predict the stimulus from the neural response. Predicting the stimulus from the response is known as neural decoding. -## Data generation -A predefined filter consisting of 3 Gaussian bumps with equal covariance (A) resembled the shape of the filter used by Sharpee et al., but here the Gaussian bumps are offset instead of being centered. This filter corresponds to the linear weights of a model simple cell, a stochastic neuron. The inner product between an input image and the filter, denoted $s$, is proportional to the probability of the neuron spiking/firing or not. Specifically, a zero-mean Gaussian random variable $e$ with variance $a$ is added to the inner-product, if this sum is greater than the threshold $b$ then a spike is generated. As input, I use patches from a database of natural images (consisting of buildings, parks, trees, etc.): +To guide the choice of metric learning, we propose to use centered alignment, a kernel-based dependence measure, to measure the correlation between the similarity in the data space (determined by the metric) and the similarity in the stimulus space. We provide mathematical formulation of the appropriate kernel and distance functions, as well as a MATLAB implementation of the batch and mini-batch optimization of the centered alignment metric learning (CAML). -Independent Component Filters of Natural Images Compared with Simple Cells in Primary Visual Cortex -J. H. v. Hateren and A. v. d. Schaaf -Proceedings: Biological Sciences 265 359-366 (1998) - -![Figure 1](/images/research/neuralDecoding1.png "Figure 1") -30 by 30 patches were randomly sampled from the images. The simulated cells parameters $a$ and $b$ are set relative to the standard deviation of $s$. Specifically $a=0.31\sigma(s)$ and $b=1.8sigma(s)$, using the same values as Sharpee et al. The absence or presence of spike for a given patch is treated as a label. 40,000 patches and the corresponding labels were given to the metric learning algorithm. Mini-batch optimization was run and the results are displayed for a Mahalanobis metric (B) and a weighted metric (C). To our knowledge, this was the first attempt to use a weighted metric algorithm to infer the importance of individual pixels on a simulated simple cell. diff --git a/content/research/news.md b/content/research/news.md new file mode 100644 index 0000000..3161f95 --- /dev/null +++ b/content/research/news.md @@ -0,0 +1,91 @@ +--- +title: "News and Recent Research" +shortTitle: "News!" +shortDescription: "" + +imageLink: "/images/evanshall.png" + + +date: 2021-08-24T00:59:28-04:00 +draft: false +--- + +##### Application Deadlines +* University of Delaware's Ph.D. program in Electrical and Computer Engineering (apply early; February 1 is the deadline to be considered for departmental funding.) +* University of Delaware's Undergraduate Summer Scholars (Please contact me ASAP as deadlines are usually Feb. or Mar.) +* University of Delaware's Latin and South America Summer Program for Visiting Researchers [2022 Summer Program Application Form](https://forms.gle/V88AeEDrZgkk2R488) Deadline: February 15th, 2022. + +##### December 2021 +—Good luck to all the conference presenters this month: +* ACNP: **Kristina Holton**, S. Y. Chan, A. J. Brockmeier, D. Öngür, and M-H. Hall "Exploring the influences of functional connectivity architecture on cortical thickness networks in patients with early psychosis", *60th Annual Meeting of the American College of Neuropsychopharmacology*, San Juan, Puerto Rico, 12/2021. +* Machine Learning for Health (ML4H): **Hassan Baker**, A. J. Brockmeier "Exploring latent networks in resting-state fMRI using voxel-to-voxel causal modeling feature selection" [Paper](https://arxiv.org/pdf/2111.07488.pdf), [Poster](/other/baker_poster_ML4H2021.pdf) +* Machine Learning for Health (ML4H): **Kristina Holton**, S. Y. Chan, A. J. Brockmeier, D. Öngür, and M-H. Hall "Exploring latent networks in resting-state fMRI using +voxel-to-voxel causal modeling feature selection" +* DistShift @ NeurIPS: **Yuksel Karahan**, **Bilal Riaz**, and A. J. Brockmeier "Kernel landmarks: An empirical statistical approach to detect covariate shift", *Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021)* [Paper](https://openreview.net/pdf?id=Wu5hMMQ76OE), [Poster](/other/karahan_poster_distshift2021), [5 min Presentation](https://recorder-v3.slideslive.com/?share=56618&s=79a6864d-0cec-41ad-a152-09cceb1c6bf4). +* DistShift @ NeurIPS: A. J. Brockmeier, **Claudio Cesar Claros Olivares**, M. S. Emigh, and L. G. Sanchez Giraldo "Identifying the instances associated with distribution shifts using the max-sliced Bures divergence", *Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021)* [Paper](https://openreview.net/pdf?id=y_s0M6OtyH_), [Poster](/other/brockmeier_poster_distshift2021), [5 min Presentation](https://recorder-v3.slideslive.com/?share=56676&s=b55c603d-1e31-4ab2-9cd7-440e5d05d64d). + +—Carlos presents his proposal "Learning Representative Waveforms to Analyze, Summarize, and Compare Long-term Neural Recordings" +##### November 2021 +—Kris wins best (joint) poster award at [DSI's 2021 Delaware Data Science Symposium](https://dsi.udel.edu/events/dsi-symposium-2021/) with her work "Exploring the influences of functional connectivity architecture on cortical thickness networks in patients with early psychosis" [Poster](https://dsi.udel.edu/files/formidable/45/kmholton.Poster-9fbc84.pdf), [Lightning presentation](https://dsi.udel.edu/files/formidable/45/kmholton.dsi-8c51f8.mp4) + +—Carlos presented his work at [EMBC'21](https://embc.embs.org/2021/), "Shift-invariant waveform learning on epileptic ECoG" [PDF](https://arxiv.org/abs/2108.03177.pdf) + +##### October 2021 +—Kristina (Kris) Holton, Certificate Program for Bioinformatics and Data Science, abstract and extended abstract accepted to Annual Meeting of the American College of Neuropsychopharmacology (ACNP) and Machine Learning for Health (ML4H). + +—Hassan's extended abstract also accepted to Machine Learning for Health (ML4H). + +—César's paper and Yüksel and Bilal's paper accepted to DistShift at NeurIPS 2021. + +##### September 2021 +—Hassan passes the Ph.D. qualifying examination. + +—Dr. Brockmeier will serve as Co-Investigator on a NIH-funded R21 project "Integrative Data Analysis to Identify Persistent Post-Concussion Deficits and Subsequent Musculoskeletal Injury Risk" led by Dr. Thomas Buckley + +—NSF funds a "Harnessing the Data Revolution" traineeship on "Computing and Data Science Training for Materials Innovation, Discovery, Analytics" [(MIDAS) ](https://sites.udel.edu/midas-nrt/) led by Dr. Arthi Jayaraman. (Dr. Brockmeier will serve as senior personnel.) + +* 1st year Ph.D. students interested in using AI, machine learning, and data science to solve problems in polymeric material design +* 2nd year Ph.D. students will have a pilot offering of the problem-focused team oriented 'hackathon' course + + — NSF funds a "Harnessing the Data Revolution" undergraduate research and training "Delaware and Mid-Atlantic Data Science Corps" [NSF 2123264](https://www.nsf.gov/awardsearch/showAward?AWD_ID=2123264) led by Dr. Federica Bianco *et al*. (Dr. Brockmeier will serve as senior personnel.) + + + +##### August 2021 +—Summer Scholars (Evan Curtin and Justin Labombard) present at the Undergraduate Research and Service Scholar Celebratory Symposium +* Evan Curtin, "Non-Negative Matrix Factorization as Dictionary Learning for Audio Separation" [Poster PDF](other/curtin_poster_summer2021.pdf) +* Justin Labombard, "Dictionary Learning on Eplieptic ECoG Waveforms" [Poster PDF](other/labombard_poster_summer2021.pdf) +—Kick off of new NSF-funded project led by Dr. Federica Bianco on **AI for astronomy**: ``Detecting and studying light echoes in the era of Rubin and artificial intelligence'' [NSF 2108841](https://nsf.gov/awardsearch/showAward?AWD_ID=2108841) + +##### July 2021 +—[Andres Nicolas Lopez](/members/#nicolas), MSc. (National University of Colombia), joins us as a visiting research working on modeling the uncertainty and errors of machine learning models [University of Delaware's Latin and South America Summer Program](https://www.eecis.udel.edu/~arce/partnerships/summerprogram) +—Carlos's second paper accepted to EMBC'21 +* "Shift-invariant waveform learning on epileptic ECoG" [Pre-print PDF](http://arxiv.org/abs/2108.03177) + +##### July 2021 +—Evan, Justin, and Thomas start the (undergraduate) Summer Scholars Program + +##### May 2021 +—Carlos and Hassan presented their work as posters at NER'21 +* Hassan presented "Local and Sparse Linear Causal Models for fMRI Resting-State Signals" [Full-text (IEEE)](https://ieeexplore.ieee.org/document/9441242/) | [Poster PDF](/other/baker_poster_NER2021.pdf) +* [Carlos](/members/#carlos) presented "Searching for Waveforms on Spatially-Filtered Epileptic ECoG" [Full-text (IEEE)](https://ieeexplore.ieee.org/document/9441220/) | [PDF](https://arxiv.org/pdf/2103.13853.pdf) | [Poster PDF](/other/carlos_poster_NER2021.pdf) +* We made the top-5 finalist for the Neural Engineering Community Award + +##### April 2021 +—Kick off of new ONR-funded project on [**interpretable divergences**](/research/discrepancies/) +* Preliminary work on [OpenReview]("https://openreview.net/pdf?id=D2Fp_qheYu") + +##### January 2021 +— [Hassan](/members/#hassan) awarded Gore Engineering Graduate Fellowship for Spring 2021. + +—[Hau Phan](/members/#hau) joins the lab to work on a collaborative ONR-funded project for bioinspired processing of chemical signals. + + +##### September 2020 +[Bilal](/members/#bilal) passes the Ph.D. qualifying examination. + +##### August 2020 +[César](/members/#cesar) completes his Master's thesis "Synergistic Human-Machine Prediction: Active Error Analysis and Mitigation with Gaussian Process Regression". + +##### February 2020 +[Hassan](/members/#hassan) joins the lab to work on a [neuroimaging and graph signal processing project](/research/graph/). \ No newline at end of file diff --git a/content/research/pico.md b/content/research/pico.md new file mode 100755 index 0000000..0f2b94d --- /dev/null +++ b/content/research/pico.md @@ -0,0 +1,43 @@ +--- +title: "Improving reference prioritisation with PICO recognition" +shortTitle: "Reference Prioritisation with PICO" +shortDescription: "Machine learning can assist with multiple tasks during systematic reviews to facilitate the rapid retrieval of relevant references during screening and to identify and extract information relevant to the study characteristics, which include the PICO elements of patient/population, intervention, comparator, and outcomes." + +authors: "Austin J. Brockmeier, Meizhi Ju, Piotr Przybyła, Sophia Ananiadou" +journal: "BMC Medical Informatics and Decision Making" +volume: 19 +issue: 1 +pages: "1–14" +year: 2019 + +imageLink: "/images/research/pico1.png" +externalLink: +journalLink: "https://doi.org/10.1186/s12911-019-0992-8" +pdfLink: "/papers/brockmeier2019_improving_reference_prioritisation.pdf" +codeLink: + +date: 2019-06-01T00:59:28-04:00 +draft: false +--- + +## Overview + +![Systematic reviews of biomedical research focus on specific questions framed in terms of the patient population, intervention, control, and outcomes measured (PICO). These terms are used to query literature databases. The resulting hits (which could be hundreds or thousands of articles) need to be sorted into relevant and irrelevant groups.](/images/research/pico2.png "Example of systematic reviews. ") + +Machine learning can assist with multiple tasks during systematic reviews to facilitate the rapid retrieval of relevant references during screening and to identify and extract information relevant to the study characteristics, which include the PICO elements of patient/population, intervention, comparator, and outcomes. + + +![Example of automatically extracted participants, intervention, and outmodes (PIO). Techniques similar to named entity recognition are used for identifying and categorising fragments of text. ](/images/research/pico3.png "Example of PICO. ") + + +A publicly available corpus of PICO annotations on biomedical abstracts is used to train a named entity recognition model, which is implemented as a recurrent neural network. This model is then applied to a separate collection of abstracts for references from systematic reviews within biomedical and health domains. The occurrences of words tagged in the context of specific PICO contexts are used as additional features for a relevancy classification model. Simulations of the machine learning-assisted screening are used to evaluate the work saved by the relevancy model with and without the PICO features. + +![PICO recognition and abstract screening process. In the first phase, the PICO recognition model is trained to predict the PICO mention spans +on a human annotated corpus of abstracts. In the second phase, a collection of abstracts is processed by the PICO recognition model and the +results along with the original abstract are used to create a vector representation of each abstract. In the final phase, a user labels abstracts as being +included (relevant) or excluded, these decisions are used to train a machine learning (ML) model that uses the vector representation. The ML model +is applied to the remaining unlabelled abstracts, which are then sorted by their predicted relevancy, the user sees the top ranked abstracts, labels +them, and this process repeats ](/images/research/pico4.png "Data flow across the entire project. ") + + + Chi-squared and statistical significance of positive predicted values are used to identify words that are more indicative of relevancy within PICO contexts. Inclusion of PICO features improves the performance metric on 15 of the 20 collections, with substantial gains on certain systematic reviews. Examples of words whose PICO context are more precise can explain this increase. Words within PICO tagged segments in abstracts are predictive features for determining inclusion. Combining PICO annotation model into the relevancy classification pipeline is a promising approach. \ No newline at end of file diff --git a/content/research/test.md b/content/research/test.md deleted file mode 100644 index 40dda3b..0000000 --- a/content/research/test.md +++ /dev/null @@ -1,22 +0,0 @@ ---- -title: "External Link" -shortTitle: "Test" -shortDescription: "Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat." - -authors: "" -journal: -volume: -issue: -pages: -year: - -imageLink: -externalLink: "www.google.com" -journalLink: -pdfLink: -codeLink: - -date: 2020-06-01T00:59:28-04:00 -draft: false ---- - diff --git a/content/singles/admissions.md b/content/singles/admissions.md old mode 100644 new mode 100755 index bad0a11..4215530 --- a/content/singles/admissions.md +++ b/content/singles/admissions.md @@ -4,17 +4,21 @@ date: 2021-04-28T12:49:16-04:00 draft: false --- -I will be accepting graduate students beginning with the 2020–2021 academic year and continuing through subsequent semesters. In exceptional situations, support may be available starting in the summer before the fall term. +I will be accepting graduate students again (after the pandemic) starting Spring 2022 or Fall 2022 and continuing through subsequent semesters. In exceptional situations, support before the fall term. -Due to the fact that I may miss your email, please apply to the [Department of Electrical and Computer Engineering (ECE)](http://ece.udel.edu/index.html) and designate me as a potential advisor. I may also be able to advise students in the [Department of Computer and Information Sciences (CIS)](http://www.cis.udel.edu/): please designate me as a potential advisor. Both departments are in the [College of Engineering](https://www.engr.udel.edu/) at the University of Delaware. +Due to the fact that I may miss your email, please apply to the [Department of Electrical and Computer Engineering (ECE)](http://ece.udel.edu/index.html) and designate me as a potential advisor. I am also able to advise students in the [Department of Computer and Information Sciences (CIS)](http://www.cis.udel.edu/). Both departments are in the [College of Engineering](https://www.engr.udel.edu/) at the University of Delaware. Starting Fall 2022, we are also specifically looking for students interested in pursuing a Ph.D. focused on data science and computing applied to polymer design and innovation. [Find out more](https://sites.udel.edu/midas-nrt/) + +Also starting in Fall 2022, UD is recruiting for its [interdisciplinary neuroscience graduate (ING) program](https://www.udel.edu/academics/colleges/grad/prospective-students/programs/interdisciplinary/interdisciplinary-neuroscience/) that spans many colleges including College of Engineering. Applicants with backgrounds in computational and analytic fields who are interested in advancing data science tools for understanding the brain are encouraged to apply. + +In CIS, the admission review is performed by a committee. Applicants are encouraged to have a focused statement describing their research pursuits and their skills and experiences. In ECE, the admission review also involves the associate chair of graduate studies. The process has many stages: Applications received by Feb. 1 or earlier will receive primary consideration for funding. The research statement is important during the evaluation. Applications are reviewed along with those for other suitable advisors. - -- (Early February) A fair assessment of the applications is made and a subset of the qualified applicants may be short-listed. -- (Feb–March) I plan to schedule and hold interviews with short-listed candidates on video teleconferencing. -- (March) The associate chair of graduate studies make the final determination to offer funding for a certain duration. -- (April 15th) Applicants should review the offer and accept it, or decline it if they choose to take another option. The deadline for accepting an offer of financial support from a graduate school in the US is April 15th. [(PDF)](/other/graduatePdf2018.pdf) +- (January) Exceptional candidates may be contacted. +- (Early February) An assessment of the applications is made and a subset of the qualified applicants may be short-listed. +- (Feb–March) I typically schedule and hold short interviews with short-listed candidates on video teleconferencing along with current members of my group. I encourage applicants to contact and discuss the research and lab culture with current students. +- (March) Depending on funding, I along with the associate chair of graduate studies make the final determination to offer funding for a certain duration. +- (April 15th) Applicants should review the offer and accept it, or decline it if they choose to take another option. The deadline for accepting an offer of financial support from a graduate school in the US is April 15th. [(PDF)](https://cgsnet.org/ckfinder/userfiles/files/CGS_April15_Resolution_Oct2020Revision.pdf) diff --git a/content/singles/contact.md b/content/singles/contact.md old mode 100644 new mode 100755 index 8db8572..fc7c007 --- a/content/singles/contact.md +++ b/content/singles/contact.md @@ -6,8 +6,8 @@ layout: "contact" email: "ajbrock@udel.edu" -roomLocation: "Evans Hall Room 306" -address: "139 The Green, Newark, DE 19716 aaaa" +roomLocation: "Evans Hall, Room 306" +address: "139 The Green, Newark, Delaware 19716" googleEmbedLink: "https://www.google.com/maps/embed?pb=!1m18!1m12!1m3!1d3070.64487299403!2d-75.75332128462634!3d39.68020237945748!2m3!1f0!2f0!3f0!3m2!1i1024!2i768!4f13.1!3m3!1m2!1s0x89c7aa2fc8419f8d%3A0x84148407eaaa0d59!2sEvans%20Hall%2C%20139%20The%20Green%20%23306%2C%20Newark%2C%20DE%2019716!5e0!3m2!1sen!2sus!4v1619648304977!5m2!1sen!2sus" diff --git a/content/singles/education.md b/content/singles/education.md old mode 100644 new mode 100755 index e6a63d4..d0d6cda --- a/content/singles/education.md +++ b/content/singles/education.md @@ -4,14 +4,102 @@ date: 2021-06-01T02:38:52-04:00 draft: false --- -## Heading 1 +#### *Fall semesters*: second-year graduate-level course in [machine learning](#eleg817). +#### *Spring semesters*: 300-level undergraduate course in [signal processing and linear system analysis](#eleg305) +#### *Summer 2021*: [Independent study on adaptive filters](#eleg466) +#### *Spring 2021*: Graduate Data Science Community Hour during [Spring 2020](https://dsi.udel.edu/events/data-science-community-hour/) -Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Vulputate eu scelerisque felis imperdiet proin fermentum leo vel orci. Donec enim diam vulputate ut pharetra sit amet aliquam id. Senectus et netus et malesuada fames ac. Nunc mattis enim ut tellus elementum sagittis. Suscipit tellus mauris a diam maecenas sed enim. Lacus vestibulum sed arcu non. Cursus mattis molestie a iaculis at erat pellentesque adipiscing. Risus ultricies tristique nulla aliquet enim tortor at auctor urna. Nunc id cursus metus aliquam eleifend mi. Faucibus vitae aliquet nec ullamcorper sit. Mauris cursus mattis molestie a iaculis at. Luctus accumsan tortor posuere ac ut consequat semper. -![Sample Image](/images/logo-black-fill.svg) -Cras sed felis eget velit aliquet. Nisi porta lorem mollis aliquam. Pharetra sit amet aliquam id diam maecenas ultricies mi eget. Feugiat pretium nibh ipsum consequat. At risus viverra adipiscing at in tellus. Vel elit scelerisque mauris pellentesque. Scelerisque in dictum non consectetur a erat nam at. Vel pretium lectus quam id leo in vitae turpis massa. Platea dictumst vestibulum rhoncus est pellentesque. Massa eget egestas purus viverra accumsan in nisl nisi scelerisque. Leo a diam sollicitudin tempor. Viverra accumsan in nisl nisi scelerisque. +## Graduate Course -## Heading 2 +### Large-Scale Machine Learning *ELEG/FSAN 817* {#eleg817} +**Course Description:** Introduction to the analysis and processing of massive high-dimensional data. Massive data sets generally involve growth not only in the number of individuals represented but also in the number of descriptive parameters of the individuals, leading to exponential growth in the number of hypothesis considered. Approaches to address these problems exploit concepts from statistics and machine learning. -Quam vulputate dignissim suspendisse in. Nascetur ridiculus mus mauris vitae. Pharetra et ultrices neque ornare. Praesent semper feugiat nibh sed pulvinar proin gravida. Aenean euismod elementum nisi quis eleifend quam. Tristique nulla aliquet enim tortor at auctor urna. Eu volutpat odio facilisis mauris sit amet massa vitae. Donec massa sapien faucibus et molestie ac. Facilisis magna etiam tempor orci. Risus sed vulputate odio ut enim blandit volutpat maecenas. Commodo ullamcorper a lacus vestibulum sed arcu non odio euismod. Aliquam sem fringilla ut morbi tincidunt. Amet venenatis urna cursus eget nunc scelerisque viverra mauris in. In egestas erat imperdiet sed euismod nisi porta lorem mollis. +**Offered:** Fall 2019, Fall 2020, [Fall 2021](https://dsi.udel.edu/files/formidable/19/Large-Scale-Machine-Learning-Syllabus-2021.pdf) + +The course serves as a capstone course in machine learning for graduate students, with the goal of readying students for machine learning and data science research. Computational and statistical scaling of machine learning is discussed from both theoretical and practical perspectives. The formative assessments provide exercise in key aspects of machine learning research: problem formulation, critical reading of literature, algorithm implementation, abstract writing, experimental design, and peer review. Summative assessment is based on a mid-term exam and final project. + +**2020 online synchronous course enrollment:** There were 17 students enrolled: ECE (9 Ph.D. / 2 M.S.), FSAN (3 Ph.D.), Computer Science (1 Ph.D.), and Physics and Astronomy (1 Ph.D.). One additional ECE Ph.D. student audited. + +**2019 course enrollment:** There were 13 students enrolled: ECE (3 Ph.D. / 2 M.S. / 1 B.S./M.S.), Computer Science (5 Ph.D.), Economics (1 Ph.D.), Civil \& Environmental Eng. (1 Ph.D., auditing). + + +**Anonymous feedback from 2020 course evaluations:** + - "The topics covered in the course are interesting and relevant. The instructor exhibits a deep understanding of the subject and has the capacity to motivate interest in the students." + - "The instructor did a great job structuring the course and selecting the key concepts and methods among the vast amount of topics in large-scale machine learning. He is very knowledgeable and demonstrates an honest interest in our learning experience. The evaluation tools used (the midterm, the pop quizzes, the homeworks, and the final project) are well designed to accomplish our learning goals. I particularly liked that he made emphasis throughout the course in that the evaluation tools were more intended as learning tools, and offer us options to ease the natural pressure that the grades associated with those tools usually have." + - "I like the discussions and that despite being online, the professor was capable of keeping me engaged even when the topics were not particularly easy to explain. I appreciate the use of a whiteboard instead of the slides, it gives flexibility to the class and it helps me to understand." + - "He always wants us understand the topics. He supports lecture[…] interactions. He asks during lectures to make sure we are on the right track. He tried to link topics that are correlated and related to each other. He links the topics we are studying to a real world examples." + - "I can see Prof. Brockmeier really wants students to learn, no matter whether a student is already a machine learning research[er] in a specific area or is still exploring the subject." + - "Dr. Brockmeier is a great professor. He was well prepared for the classes and was capable of address almost every question that pops up during the course and always made a good connection between theory and Matlab examples." + - "There will be timely comments and replies to every student’s questions and papers" + - "He was open to listening to the questions of all the students that wanted to ask and address those questions in a respectful way." + - "He was always polite and gave everyone the opportunity to express their opinions and ideas regarding the topics that were explained." + - "I rank him in the top 1 percentile for research ability, approachability and overall being a good person. I would take his class again and recommend it people." + + +**Anonymous feedback from 2019 course evaluations:** + - "The course really dives deep into the mathematics and theory of the machine learning techniques. The organization of the course was nice." + - "Course material was well ordered, well motivated, and useful." + - "The breadth of topics studied in this class provided me with a better understanding of machine learning in general." + - "Professor Brockmeier [...] was able to connect abstract concepts from the class to multiple applications to motivate understanding" + - "This was not only the best course I took as a graduate student, but moreover, the most useful in the pursuit of becoming a machine learning expert." + - "He was enthusiastic about delivering the contents of the course." +## Undergraduate Courses + +### Signals and Systems *ELEG 305* {#eleg305} +This course examines continuous and discrete-time signals and systems at the introductory +level. + +**Course description:** Introduction to signals and systems, with an emphasis on time and frequency characterization of linear, time-invariant systems. Covers discrete and continuous time systems, sampling, and Fourier, Laplace, and Z transforms. + +**Offered:** Spring 2020, +[Spring 2021](https://dsi.udel.edu/files/formidable/19/ELEG305-S21.pdf), Spring 2022 + + +*Core requirement in the following programs:* +- Electrical Engineering +- Computer Engineering + and computer engineering majors. +- GIScience and Environmental Data Analytics + +*Technical elective:* +- Chemical Engineering, Materials Science and Engineering, Mechanical Engineering + +*Prerequisite for courses:* +- ELEG 306 - Digital Signal Processing +- ELEG 403 - Communication Systems Engineering +- ELEG 404 - Digital Imaging and Photography +- ELEG 418 - Digital Control Systems +- ELEG 492 - Radar Systems and Technology +- BMEG 441 - Biomechatronics† +- BMEG 330 - Biomedical Instrumentation† +† As an alternative for BMEG 230 + + +**Honors Discussion Section:** +Students registered in Honors for this class will be involved in an additional discussion section, which introduces content beyond the general class: specifically, non-linear, adaptive, and random systems. They will also be required to complete a semester long project that reflects an application of such systems. + +**Unsolicited Feedback from 2021 Honors section:** +- "I really enjoyed looking more into something I’m very interested in. It was my favorite Zoom meeting between all of my classes this semester. Connecting with what I focused on, [...] Doing Honors research got me to realize that dream." + +**Unsolicited Feedback from 2020 section:** + +- “I want to applaud your tenacity and optimism throughout the term. Despite this class being the most challenging course I've ever taken, I have enjoyed being your student. + + +### Independent Study *ELEG 466* {#eleg466} +"Adaptive Filtering as an Introduction to Machine Learning" + +**Offered:** Summer 2021 + +**Course description:** Introduction to adaptive filter theory and application. Students will create a technical report that details the theory, implementation, and computational experiments with the application of adaptive filtering. Topics to be covered include review of delay lines, z-transform, multivariate signal processing, Gaussian white noise, correlation matrices, whitening transforms, least squares, convex optimization, Wiener filter, stochastic gradient descent, least mean squares (LMS), NLMS, common spatial patterns, and various types of recurrent memory. A comprehensive technical report with equations and figures will be typeset in Latex. The report will detail the problem formulations, equations, and vector graphics of block diagrams, plots of results. Software implementations will be made in MATLAB/Octave, Julia, python/Numpy, or python/Pytorch. Synthetic experiments will be conducted for noise cancellation, channel estimation, and machine learning. + +**Learning outcomes:** +While students typically learn theory and fundamentals of signal processing and filtering in ELEG305 “Signals and Systems” and probability theory and statistics related to signals in ELEG310, due to time constraints and curriculum design, students rarely have the opportunity to learn and apply adaptive filter theory that combines filtering, linear algebra, optimization, and statistics together. This combination is the foundation of machine learning theory and algorithms. This independent study will attempt to fill the gap in the current curriculum areas in theory and practice. A combination of synthetic data and real-world projects will be explored by students to master these concepts. Students will practice their understanding by describing the formulations and results (equation and paper typesetting, block diagrams, data visualization and presentation) in their own words as well as implementing and testing software for adaptive filtering and machine learning. In summary the students will learn the foundations and techniques of machine learning: linear and non-linear filtering, supervised learning, least squares, optimization, stochastic gradient descent, batch normalization, and experimental design (training, validation, and testing). + +**Experience:** +The course will solidify knowledge learned in other courses and provide the student with a diverse and broad understanding of adaptive filtering, and an introduction to machine learning. The course project will require tools learned in other academic classes (signal and systems, linear algebra, calculus, probability, and statistics) and will help the student in future classes (communication systems, machine learning, control theory) as well as research. + +**Expectations:** +Students will be required to complete a technical report which will be a culmination of most of the topics covered in class. Weekly revised versions of the technical report must be submitted to document learning progress until the end of the course. Each of these weekly submissions, will require an executive summary that overviews the specific learning activities, progress in understanding and experiments, obstacles, and interesting insights or useful skills that have been learned. Students can work in groups and collaborate in a civil and equal manner. \ No newline at end of file diff --git a/content/singles/outreach.md b/content/singles/outreach.md old mode 100644 new mode 100755 index be46bd7..ab5c8c0 --- a/content/singles/outreach.md +++ b/content/singles/outreach.md @@ -5,14 +5,17 @@ draft: false --- ![Outreach Image](/images/outreach/image4.jpg) -In 2019, I took part in a Women in Engineering organized outreach event at a local New Castle Middle School, [Serviam Girls Academy](https://serviamgirlsacademy.org/) , whose mission is to take girls from low-income houses and give them the education they need to succeed in high school and beyond. For this purpose, I developed a module showing ECE’s critical role in neuroimaging technology and analysis for their enrichment program during their enrichment hour (4-5 pm on Wednesday). Working with ECE undergraduate student (Trevor Foresta) and a post-doctoral research Julia Schneider from the Department of Linguistics and Cognitive Sciences. Trevor assembled and tested the OpenBCI headset. +In 2019, we joined UD's Women in Engineering outreach program at a local New Castle Middle School, [Serviam Girls Academy](https://serviamgirlsacademy.org/) , whose mission is to take girls from low-income houses and give them the education they need to succeed in high school and beyond. Working with Trevor Foresta, ECE undergraduate student and Julia Schneider, a post-doctoral researcher from the Department of Linguistics and Cognitive Sciences, we developed a module showing ECE’s critical role in neuroimaging technology and analysis for their enrichment program during their enrichment hour (4-5 pm on Wednesday). Julia provided slides and enthusiasm on current and past sensors and systems to measure the brains operation. Trevor assembled and tested the OpenBCI headset while still a first-year engineering student. ![Outreach Image](/images/outreach/image8.jpg) ![Outreach Image](/images/outreach/image5.jpg) ![Outreach Image](/images/outreach/image7.jpg) -In 2020, I volunteered and hosted a lunchtime activity booth at the “Engineering Your Tomorrow” event held at Delaware Tech Community College in Georgetown, DE. The Sussex County STEM Alliance organized this outreach event with the purpose “to expose, mentor and teach our County’s diverse young girls the basic fundamentals of science and engineering through high interest STEM activities, while showing them... "STEM IS FUN". After helping with the entomology experiments in the morning, I set up the booth that showcased ECE’s critical role in biomedical engineering with a live demonstration and interactive handouts. For this event, I worked with ECE undergraduate student (Trevor Foresta). Trevor had previously assembled and tested the OpenBCI headset. +In 2020, Trevor and I (along with other volunteers) hosted a lunchtime activity booth at the “Engineering Your Tomorrow” event held at Delaware Tech Community College in Georgetown, DE. The Sussex County STEM Alliance organized this outreach event with the purpose “to expose, mentor and teach our County’s diverse young girls the basic fundamentals of science and engineering through high interest STEM activities, while showing them... "STEM IS FUN". After helping with the entomology experiments in the morning, I set up the booth that showcased ECE’s critical role in biomedical engineering with a live demonstration and interactive handouts. Thanks to UD's Erin Sparks for inviting us to this event. ![Outreach Image](/images/outreach/image1.jpg) ![Outreach Image](/images/outreach/image3.jpg) ![Outreach Image](/images/outreach/image6.jpg) + +{{< figure src="/images/outreach/image0.png" title="Neural Engineering Community Award" class="center" width="450">}} +Our initial outreach efforts were recognized by being one of five short-listed groups for the Neural Engineering Community Award at the 2021 Neural Engineering Workshop (NER'21)*, a biennial conference hosted by the IEEE Engineering in Medicine & Biology Society. The award was for outstanding contributions to the mission of inclusion and diversity. Congratulations to the winners! We will increase our impact and efforts in 2021–2022. \ No newline at end of file diff --git a/development_notes.txt b/development_notes.txt new file mode 100644 index 0000000..d80fdcc --- /dev/null +++ b/development_notes.txt @@ -0,0 +1,12 @@ +Education: +Add links to technical prerequisite interviews. + +Outreach: +Add link to Erin Sparks's UD webpage. + +Make PDF of outreach flyer (a JPEG thumbnail) linked to clicking on image. Overall make thumbnails clickable. + + + + + diff --git a/layouts/.DS_Store b/layouts/.DS_Store new file mode 100644 index 0000000..6056f3a Binary files /dev/null and b/layouts/.DS_Store differ diff --git a/layouts/_default/.DS_Store b/layouts/_default/.DS_Store new file mode 100644 index 0000000..ca6af2d Binary files /dev/null and b/layouts/_default/.DS_Store differ diff --git a/layouts/_default/.index.html.swp b/layouts/_default/.index.html.swp new file mode 100644 index 0000000..b9f14ea Binary files /dev/null and b/layouts/_default/.index.html.swp differ diff --git a/layouts/_default/_markup/render-image.html b/layouts/_default/_markup/render-image.html old mode 100644 new mode 100755 diff --git a/layouts/_default/baseof.html b/layouts/_default/baseof.html old mode 100644 new mode 100755 diff --git a/layouts/_default/index.html b/layouts/_default/index.html old mode 100644 new mode 100755 index 2398c24..4598be0 --- a/layouts/_default/index.html +++ b/layouts/_default/index.html @@ -1,15 +1,24 @@ {{ define "main" }}

Computational Neural Information Engineering Lab

- +
-

Welcome to CNI Engineering Lab! CNI Engineering is a computational and statistical neuroscience lab based at the University of Delaware. The primary goals of our lab include developing new techniques for processing and analyzying data from the central or peripheral nervous system using electrophysiological recordings (EEG, EMG, ECoG) and other brain imaging technology (fMRI, NIRS).

+

Welcome to the CNIEL website!

+ + +

CNIEL is lead by Austin J. Brockmeier, Ph.D.

+

Assistant Professor in the Department of Electrical and Computer Engineering (ECE) and the Department of Computer and Information Sciences (CIS) within the College of Engineering at the University of Delaware (UD), and a resident faculty in UD's Data Science Institute.

+Email | Google Scholar +

+

+

CNIEL is a team of researchers: computer scientists and engineers (electrical, computer, and neural) engaged in research, education, and outreach, focused on Data Science, Machine Learning, and Signal Processing in order to extract actionable information from complex data, especially to understand and interface with the brain. Please check out the news!

+ + + + + + + + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/index.xml b/public/index.xml new file mode 100644 index 0000000..9489594 --- /dev/null +++ b/public/index.xml @@ -0,0 +1,220 @@ + + + + CNIEL - University of Delaware + https://www.eecis.udel.edu/~ajbrock/ + Recent content on CNIEL - University of Delaware + Hugo -- gohugo.io + en-us + Tue, 24 Aug 2021 00:59:28 -0400 + + News and Recent Research + https://www.eecis.udel.edu/~ajbrock/research/news/ + Tue, 24 Aug 2021 00:59:28 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/news/ + Application Deadlines University of Delaware&rsquo;s Ph.D. program in Electrical and Computer Engineering (apply early; February 1 is the deadline to be considered for departmental funding.) University of Delaware&rsquo;s Undergraduate Summer Scholars (Please contact me ASAP as deadlines are usually Feb. or Mar.) University of Delaware&rsquo;s Latin and South America Summer Program for Visiting Researchers 2022 Summer Program Application Form Deadline: February 15th, 2022. December 2021 —Good luck to all the conference presenters this month: + + + + Visiting Scholar (Summer 2021) + https://www.eecis.udel.edu/~ajbrock/members/nicolaslopez/ + Sun, 20 Jun 2021 17:36:38 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/nicolaslopez/ + Statistician working on the error modeling for synergistic machine learning. LinkedIn + + + + Undergraduate (Summer Scholar 2021) + https://www.eecis.udel.edu/~ajbrock/members/justinlabombard/ + Tue, 15 Jun 2021 17:00:22 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/justinlabombard/ + Dictionary Learning on Epileptic ECoG Waveforms + + + + Undergraduate (Summer Scholar 2021) + https://www.eecis.udel.edu/~ajbrock/members/evancurtin/ + Tue, 01 Jun 2021 21:00:00 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/evancurtin/ + Non-Negative Matrix Factorization as Dictionary Learning for Audio Separation + + + + Education + https://www.eecis.udel.edu/~ajbrock/singles/education/ + Tue, 01 Jun 2021 02:38:52 -0400 + + https://www.eecis.udel.edu/~ajbrock/singles/education/ + Fall semesters: second-year graduate-level course in machine learning. Spring semesters: 300-level undergraduate course in signal processing and linear system analysis Summer 2021: Independent study on adaptive filters Spring 2021: Graduate Data Science Community Hour during Spring 2020 Graduate Course Large-Scale Machine Learning ELEG/FSAN 817 Course Description: Introduction to the analysis and processing of massive high-dimensional data. Massive data sets generally involve growth not only in the number of individuals represented but also in the number of descriptive parameters of the individuals, leading to exponential growth in the number of hypothesis considered. + + + + K–12 Outreach and Undergraduate Research + https://www.eecis.udel.edu/~ajbrock/singles/outreach/ + Thu, 29 Apr 2021 00:49:51 -0400 + + https://www.eecis.udel.edu/~ajbrock/singles/outreach/ + In 2019, we joined UD&rsquo;s Women in Engineering outreach program at a local New Castle Middle School, Serviam Girls Academy , whose mission is to take girls from low-income houses and give them the education they need to succeed in high school and beyond. Working with Trevor Foresta, ECE undergraduate student and Julia Schneider, a post-doctoral researcher from the Department of Linguistics and Cognitive Sciences, we developed a module showing ECE’s critical role in neuroimaging technology and analysis for their enrichment program during their enrichment hour (4-5 pm on Wednesday). + + + + Admissions + https://www.eecis.udel.edu/~ajbrock/singles/admissions/ + Wed, 28 Apr 2021 12:49:16 -0400 + + https://www.eecis.udel.edu/~ajbrock/singles/admissions/ + I will be accepting graduate students again (after the pandemic) starting Spring 2022 or Fall 2022 and continuing through subsequent semesters. In exceptional situations, support before the fall term. +Due to the fact that I may miss your email, please apply to the Department of Electrical and Computer Engineering (ECE) and designate me as a potential advisor. I am also able to advise students in the Department of Computer and Information Sciences (CIS). + + + + Contact + https://www.eecis.udel.edu/~ajbrock/singles/contact/ + Wed, 28 Apr 2021 12:49:01 -0400 + + https://www.eecis.udel.edu/~ajbrock/singles/contact/ + + + + + Interpretable maximal discrepancies metrics for analyzing and improving generative models + https://www.eecis.udel.edu/~ajbrock/research/discrepancies/ + Wed, 14 Apr 2021 00:59:28 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/discrepancies/ + Overview Divergence measures quantify the dissimilarity, including the distance, between distributions and are fundamental to hypothesis testing, information theory, and the estimation and criticism of statistical models. Recently, there has been renewed interest in divergences in the context of generative adversarial neural networks (GANs). While a multitude of divergences exist, they vary in their characteristics. Importantly, not all divergences are equally interpretable: a divergence between samples is considered interpretable if it directly answers the question “Which instances best exhibit the discrepancy between the samples? + + + + M.S. and Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/hauphan/ + Fri, 01 Jan 2021 17:00:52 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/hauphan/ + Reinforcement learning estimation and detection. + + + + Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/hassanbaker/ + Sat, 01 Feb 2020 17:00:38 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/hassanbaker/ + Graph signal processing for functional and structural neuroimaging. + + + + Advancing machine learning for neuroimaging through topology-aware signal processing + https://www.eecis.udel.edu/~ajbrock/research/graphs/ + Sat, 01 Feb 2020 00:59:28 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/graphs/ + The proposed work plan is to develop machine learning techniques to work directly with graph signal processing techniques in the context of neuroimaging. The goal is to leverage information in the form of the topology of the signal sensors or measurement locations to refine the neural signal representations in order to improve the statistical power of tests for distinguishing differences between conditions or stimuli. The project’s scope includes the formulation, mathematical and statistical analysis, and initial validation of the proposed methodology. + + + + Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/bilalriaz/ + Wed, 28 Aug 2019 17:00:22 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/bilalriaz/ + Unsupervised learning as optimizations with positive semidefinite matrices. + + + + Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/yukselkarahan/ + Wed, 28 Aug 2019 16:59:19 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/yukselkarahan/ + Data science techniques for semisupervised domain transfer. + + + + Visiting Scholar (Summer 2019) + https://www.eecis.udel.edu/~ajbrock/members/edwinsalcedo/ + Mon, 01 Jul 2019 17:36:38 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/edwinsalcedo/ + Machine learning for semisupervised domain transfer. + + + + Improving reference prioritisation with PICO recognition + https://www.eecis.udel.edu/~ajbrock/research/pico/ + Sat, 01 Jun 2019 00:59:28 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/pico/ + Overview +Machine learning can assist with multiple tasks during systematic reviews to facilitate the rapid retrieval of relevant references during screening and to identify and extract information relevant to the study characteristics, which include the PICO elements of patient/population, intervention, comparator, and outcomes. + +A publicly available corpus of PICO annotations on biomedical abstracts is used to train a named entity recognition model, which is implemented as a recurrent neural network. + + + + alumnus and admitted Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/cesarolivares/ + Wed, 01 May 2019 17:00:02 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/cesarolivares/ + Error-aware and vigilant machine learning. 3D convolutional neural networks for brain age prediction from structural measures. Optimal transport for color and style transfer. + + + + Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/carloscardenas/ + Tue, 01 Jan 2019 13:12:57 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/carloscardenas/ + Research focus: finding patterns in neural time series through convolutional sparse analysis. +Goal: to discover physiologically meaningful waveforms in multi-day continuous epileptic electrocorticographic (ECoG) recordings that can be used to build interpretable features for seizure prediction. +Methods: interpretable machine learning, clustering and sparse coding for time series, supervised learning for neural data + + + + Assistant Professor + https://www.eecis.udel.edu/~ajbrock/members/austinbrockmeier/ + Sat, 28 Apr 2018 17:07:03 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/austinbrockmeier/ + Curriculum Vitae Email +What: Data Science, Machine Learning, and Signal Processing +These involve the underlying mathematical analysis, design of statistical models, and software implementation of data and signal processing (filtering and neural networks) as well as optimization routines. + Finding discrepancies in data sets using divergences Selecting representations of data using informativeness measures Why: To extract actionable information from complex data, especially to understand and interface with the brain + + + + Quantifying the informativeness of similarity measurements + https://www.eecis.udel.edu/~ajbrock/research/informativeness/ + Fri, 17 Mar 2017 21:25:29 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/informativeness/ + Overview Choosing the particulars of a data representation is crucial for the successful application of machine learning techniques. In the unsupervised case, there is a lack of measures that can be used to compare different parameter choices that affect the representation. In this paper, we describe an unsupervised measure for quantifying the &lsquo;informativeness&rsquo; of correlation matrices formed from the pairwise similarities or relationships among data instances. +The measure quantifies the heterogeneity of the correlations and is defined as the distance between a correlation matrix and the nearest correlation matrix with constant off-diagonal entries. + + + + Learning Recurrent Waveforms within EEGs + https://www.eecis.udel.edu/~ajbrock/research/eegs/ + Thu, 28 Apr 2016 22:42:02 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/eegs/ + Overview When experts analyze EEGs they look for landmarks in the traces corresponding to established patterns such as oscillatory and phasic events of particular frequency or morphology. Long records motivate automated analysis techniques. Automation techniques often require design choices such as wavelet family or number of bandpass filters. To overcome this, we explore a modeling approach that automatically learns recurrent temporal waveforms within EEG traces. The estimation is based on a multiple-input, single-output linear model with sparsely excited inputs. + + + + Neural Decoding with Kernel-based Metric Learning + https://www.eecis.udel.edu/~ajbrock/research/neural_decoding/ + Mon, 28 Apr 2014 23:19:06 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/neural_decoding/ + Overview Given a sample of data points, we often assume the points reside in some space in which we can measure distances between pairs of points. From these measurements we can understand which points are close to each other. Nearby points are often assumed to share characteristics. This assumption is the foundation of nearest-neighbor classification and regression as well as clustering. +The function that measures the distance between pairs of points in a particular space is called a distance metric. + + + + diff --git a/public/members/austinbrockmeier/index.html b/public/members/austinbrockmeier/index.html new file mode 100644 index 0000000..6b16346 --- /dev/null +++ b/public/members/austinbrockmeier/index.html @@ -0,0 +1,244 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Assistant Professor

+

Curriculum Vitae +Email

+

What: Data Science, Machine Learning, and Signal Processing
+These involve the underlying mathematical analysis, design of statistical models, and software implementation of data and signal processing (filtering and neural networks) as well as optimization routines.

+ +

Why: To extract actionable information from complex data, especially to understand and interface with the brain

+ +

Google Scholar

+

Where: University of Delaware

+ + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/bilalriaz/index.html b/public/members/bilalriaz/index.html new file mode 100644 index 0000000..b40426d --- /dev/null +++ b/public/members/bilalriaz/index.html @@ -0,0 +1,223 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Ph.D. student

+
Unsupervised learning as optimizations with positive semidefinite matrices.
+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/carloscardenas/index.html b/public/members/carloscardenas/index.html new file mode 100644 index 0000000..90da6f2 --- /dev/null +++ b/public/members/carloscardenas/index.html @@ -0,0 +1,225 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Ph.D. student

+

Research focus: finding patterns in neural time series through convolutional sparse analysis.

+

Goal: to discover physiologically meaningful waveforms in multi-day continuous epileptic electrocorticographic (ECoG) recordings that can be used to build interpretable features for seizure prediction.

+

Methods: interpretable machine learning, clustering and sparse coding for time series, supervised learning for neural data

+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/cesarolivares/index.html b/public/members/cesarolivares/index.html new file mode 100644 index 0000000..2c5c48e --- /dev/null +++ b/public/members/cesarolivares/index.html @@ -0,0 +1,224 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

alumnus and admitted Ph.D. student

+
Error-aware and vigilant machine learning.
+

3D convolutional neural networks for brain age prediction from structural measures. Optimal transport for color and style transfer.

+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/edwinsalcedo/index.html b/public/members/edwinsalcedo/index.html new file mode 100644 index 0000000..86f8cd6 --- /dev/null +++ b/public/members/edwinsalcedo/index.html @@ -0,0 +1,223 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Visiting Scholar (Summer 2019)

+

Machine learning for semisupervised domain transfer.

+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/evancurtin/index.html b/public/members/evancurtin/index.html new file mode 100644 index 0000000..f326f17 --- /dev/null +++ b/public/members/evancurtin/index.html @@ -0,0 +1,223 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Undergraduate (Summer Scholar 2021)

+
Non-Negative Matrix Factorization as Dictionary Learning for Audio Separation
+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/hassanbaker/index.html b/public/members/hassanbaker/index.html new file mode 100644 index 0000000..99bc59c --- /dev/null +++ b/public/members/hassanbaker/index.html @@ -0,0 +1,223 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Ph.D. student

+
Graph signal processing for functional and structural neuroimaging.
+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/hauphan/index.html b/public/members/hauphan/index.html new file mode 100644 index 0000000..5f10b27 --- /dev/null +++ b/public/members/hauphan/index.html @@ -0,0 +1,223 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

M.S. and Ph.D. student

+
Reinforcement learning estimation and detection.
+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/index.html b/public/members/index.html new file mode 100644 index 0000000..eccedb7 --- /dev/null +++ b/public/members/index.html @@ -0,0 +1,420 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Current Members

+ +
+
+ + Austin J. Brockmeier, Ph.D. + +
+
+

Austin J. Brockmeier, Ph.D.

+
Assistant Professor
+

Curriculum Vitae +Email

+

What: Data Science, Machine Learning, and Signal Processing
+These involve the underlying mathematical analysis, design of statistical models, and software implementation of data and signal processing (filtering and neural networks) as well as optimization routines.

+ +

Why: To extract actionable information from complex data, especially to understand and interface with the brain

+ +

Google Scholar

+

Where: University of Delaware

+ + +
+
+ +
+
+ + Carlos H. Mendoza Cárdenas, M.S. + +
+
+

Carlos H. Mendoza Cárdenas, M.S.

+
Ph.D. student
+

Research focus: finding patterns in neural time series through convolutional sparse analysis.

+

Goal: to discover physiologically meaningful waveforms in multi-day continuous epileptic electrocorticographic (ECoG) recordings that can be used to build interpretable features for seizure prediction.

+

Methods: interpretable machine learning, clustering and sparse coding for time series, supervised learning for neural data

+ +
+
+ +
+
+ + C. Cesar Claros Olivares, M.S. + +
+
+

C. Cesar Claros Olivares, M.S.

+
alumnus and admitted Ph.D. student
+
Error-aware and vigilant machine learning.
+

3D convolutional neural networks for brain age prediction from structural measures. Optimal transport for color and style transfer.

+ +
+
+ +
+
+ + Yuksel Karahan, M.S. + +
+
+

Yuksel Karahan, M.S.

+
Ph.D. student
+

Data science techniques for semisupervised domain transfer.

+ +
+
+ +
+
+ + Bilal Riaz, M.S. + +
+
+

Bilal Riaz, M.S.

+
Ph.D. student
+
Unsupervised learning as optimizations with positive semidefinite matrices.
+ +
+
+ +
+
+ + Hassan Baker, M.S. + +
+
+

Hassan Baker, M.S.

+
Ph.D. student
+
Graph signal processing for functional and structural neuroimaging.
+ +
+
+ +
+
+ + Hau Phan + +
+
+

Hau Phan

+
M.S. and Ph.D. student
+
Reinforcement learning estimation and detection.
+ +
+
+ +
+
+ + Evan Curtin + +
+
+

Evan Curtin

+
Undergraduate (Summer Scholar 2021)
+
Non-Negative Matrix Factorization as Dictionary Learning for Audio Separation
+ +
+
+ +
+
+ + Justin Labombard + +
+
+

Justin Labombard

+
Undergraduate (Summer Scholar 2021)
+
Dictionary Learning on Epileptic ECoG Waveforms
+ +
+
+ + +

Past Members

+ + +
+
+ + Andres Nicolas Lopez, MSc. + +
+
+

Andres Nicolas Lopez, MSc.

+
Visiting Scholar (Summer 2021)
+

Statistician working on the error modeling for synergistic machine learning. LinkedIn

+ +
+
+ +
+
+ + Edwin Salcedo, M.Sc., M.B.A. + +
+
+

Edwin Salcedo, M.Sc., M.B.A.

+
Visiting Scholar (Summer 2019)
+

Machine learning for semisupervised domain transfer.

+ +
+
+ + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/index.xml b/public/members/index.xml new file mode 100644 index 0000000..81b74f7 --- /dev/null +++ b/public/members/index.xml @@ -0,0 +1,115 @@ + + + + Members on CNIEL - University of Delaware + https://www.eecis.udel.edu/~ajbrock/members/ + Recent content in Members on CNIEL - University of Delaware + Hugo -- gohugo.io + en-us + Sun, 20 Jun 2021 17:36:38 -0400 + + Visiting Scholar (Summer 2021) + https://www.eecis.udel.edu/~ajbrock/members/nicolaslopez/ + Sun, 20 Jun 2021 17:36:38 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/nicolaslopez/ + Statistician working on the error modeling for synergistic machine learning. LinkedIn + + + + Undergraduate (Summer Scholar 2021) + https://www.eecis.udel.edu/~ajbrock/members/justinlabombard/ + Tue, 15 Jun 2021 17:00:22 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/justinlabombard/ + Dictionary Learning on Epileptic ECoG Waveforms + + + + Undergraduate (Summer Scholar 2021) + https://www.eecis.udel.edu/~ajbrock/members/evancurtin/ + Tue, 01 Jun 2021 21:00:00 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/evancurtin/ + Non-Negative Matrix Factorization as Dictionary Learning for Audio Separation + + + + M.S. and Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/hauphan/ + Fri, 01 Jan 2021 17:00:52 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/hauphan/ + Reinforcement learning estimation and detection. + + + + Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/hassanbaker/ + Sat, 01 Feb 2020 17:00:38 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/hassanbaker/ + Graph signal processing for functional and structural neuroimaging. + + + + Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/bilalriaz/ + Wed, 28 Aug 2019 17:00:22 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/bilalriaz/ + Unsupervised learning as optimizations with positive semidefinite matrices. + + + + Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/yukselkarahan/ + Wed, 28 Aug 2019 16:59:19 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/yukselkarahan/ + Data science techniques for semisupervised domain transfer. + + + + Visiting Scholar (Summer 2019) + https://www.eecis.udel.edu/~ajbrock/members/edwinsalcedo/ + Mon, 01 Jul 2019 17:36:38 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/edwinsalcedo/ + Machine learning for semisupervised domain transfer. + + + + alumnus and admitted Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/cesarolivares/ + Wed, 01 May 2019 17:00:02 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/cesarolivares/ + Error-aware and vigilant machine learning. 3D convolutional neural networks for brain age prediction from structural measures. Optimal transport for color and style transfer. + + + + Ph.D. student + https://www.eecis.udel.edu/~ajbrock/members/carloscardenas/ + Tue, 01 Jan 2019 13:12:57 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/carloscardenas/ + Research focus: finding patterns in neural time series through convolutional sparse analysis. +Goal: to discover physiologically meaningful waveforms in multi-day continuous epileptic electrocorticographic (ECoG) recordings that can be used to build interpretable features for seizure prediction. +Methods: interpretable machine learning, clustering and sparse coding for time series, supervised learning for neural data + + + + Assistant Professor + https://www.eecis.udel.edu/~ajbrock/members/austinbrockmeier/ + Sat, 28 Apr 2018 17:07:03 -0400 + + https://www.eecis.udel.edu/~ajbrock/members/austinbrockmeier/ + Curriculum Vitae Email +What: Data Science, Machine Learning, and Signal Processing +These involve the underlying mathematical analysis, design of statistical models, and software implementation of data and signal processing (filtering and neural networks) as well as optimization routines. + Finding discrepancies in data sets using divergences Selecting representations of data using informativeness measures Why: To extract actionable information from complex data, especially to understand and interface with the brain + + + + diff --git a/public/members/justinlabombard/index.html b/public/members/justinlabombard/index.html new file mode 100644 index 0000000..94a37fe --- /dev/null +++ b/public/members/justinlabombard/index.html @@ -0,0 +1,223 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Undergraduate (Summer Scholar 2021)

+
Dictionary Learning on Epileptic ECoG Waveforms
+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/nicolaslopez/index.html b/public/members/nicolaslopez/index.html new file mode 100644 index 0000000..53febec --- /dev/null +++ b/public/members/nicolaslopez/index.html @@ -0,0 +1,223 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Visiting Scholar (Summer 2021)

+

Statistician working on the error modeling for synergistic machine learning. LinkedIn

+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/members/yukselkarahan/index.html b/public/members/yukselkarahan/index.html new file mode 100644 index 0000000..4bc644c --- /dev/null +++ b/public/members/yukselkarahan/index.html @@ -0,0 +1,223 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Ph.D. student

+

Data science techniques for semisupervised domain transfer.

+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/other/ajbrockmeier_CV_12_2021.pdf b/public/other/ajbrockmeier_CV_12_2021.pdf new file mode 100644 index 0000000..21d44b4 Binary files /dev/null and b/public/other/ajbrockmeier_CV_12_2021.pdf differ diff --git a/public/other/ajbrockmeier_CV_8_2021.pdf b/public/other/ajbrockmeier_CV_8_2021.pdf new file mode 100644 index 0000000..2e8709d Binary files /dev/null and b/public/other/ajbrockmeier_CV_8_2021.pdf differ diff --git a/public/other/baker_poster_ML4H2021.pdf b/public/other/baker_poster_ML4H2021.pdf new file mode 100644 index 0000000..e887e81 Binary files /dev/null and b/public/other/baker_poster_ML4H2021.pdf differ diff --git a/public/other/baker_poster_NER2021.pdf b/public/other/baker_poster_NER2021.pdf new file mode 100644 index 0000000..b491033 Binary files /dev/null and b/public/other/baker_poster_NER2021.pdf differ diff --git a/public/other/brockmeier_poster_distshift2021.pdf b/public/other/brockmeier_poster_distshift2021.pdf new file mode 100644 index 0000000..01b5c0e Binary files /dev/null and b/public/other/brockmeier_poster_distshift2021.pdf differ diff --git a/public/other/carlos_poster_NER2021.pdf b/public/other/carlos_poster_NER2021.pdf new file mode 100644 index 0000000..6074b73 Binary files /dev/null and b/public/other/carlos_poster_NER2021.pdf differ diff --git a/public/other/curtin_poster_summer2021.pdf b/public/other/curtin_poster_summer2021.pdf new file mode 100644 index 0000000..9b19df0 Binary files /dev/null and b/public/other/curtin_poster_summer2021.pdf differ diff --git a/public/other/graduatePdf2018.pdf b/public/other/graduatePdf2018.pdf new file mode 100755 index 0000000..794bbcf Binary files /dev/null and b/public/other/graduatePdf2018.pdf differ diff --git a/public/other/karahan_poster_distshift2021.pdf b/public/other/karahan_poster_distshift2021.pdf new file mode 100644 index 0000000..3f1dd4b Binary files /dev/null and b/public/other/karahan_poster_distshift2021.pdf differ diff --git a/public/other/labombard_poster_summer2021.pdf b/public/other/labombard_poster_summer2021.pdf new file mode 100644 index 0000000..f1ca51a Binary files /dev/null and b/public/other/labombard_poster_summer2021.pdf differ diff --git a/public/papers/brockmeier2014_kernel_metric_learning.pdf b/public/papers/brockmeier2014_kernel_metric_learning.pdf new file mode 100644 index 0000000..c606167 Binary files /dev/null and b/public/papers/brockmeier2014_kernel_metric_learning.pdf differ diff --git a/public/papers/brockmeier2016_learning_recurrent_waveforms.pdf b/public/papers/brockmeier2016_learning_recurrent_waveforms.pdf new file mode 100644 index 0000000..0ee7695 Binary files /dev/null and b/public/papers/brockmeier2016_learning_recurrent_waveforms.pdf differ diff --git a/public/papers/brockmeier2019_improving_reference_prioritisation.pdf b/public/papers/brockmeier2019_improving_reference_prioritisation.pdf new file mode 100644 index 0000000..8db87ec Binary files /dev/null and b/public/papers/brockmeier2019_improving_reference_prioritisation.pdf differ diff --git a/public/papers/brockmeier2021_ML_for_documents.pdf b/public/papers/brockmeier2021_ML_for_documents.pdf new file mode 100644 index 0000000..d3856b6 Binary files /dev/null and b/public/papers/brockmeier2021_ML_for_documents.pdf differ diff --git a/public/research/discrepancies/index.html b/public/research/discrepancies/index.html new file mode 100644 index 0000000..7565a07 --- /dev/null +++ b/public/research/discrepancies/index.html @@ -0,0 +1,248 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Interpretable Maximal Discrepancies Metrics for Analyzing and Improving Generative Models

+

+ Office of Naval Research, Grant # N00014-21-1-2300, Principal Investigator: Austin J. Brockmeier, 4/2021–4/2024.
+ + + + + +

+ + + + + + + + +
+

Overview

+

Divergence measures quantify the dissimilarity, including the distance, between distributions and are fundamental to hypothesis testing, information theory, and the estimation and criticism of statistical models. Recently, there has been renewed interest in divergences in the context of generative adversarial neural networks (GANs). While a multitude of divergences exist, they vary in their characteristics. Importantly, not all divergences are equally interpretable: a divergence between samples is considered interpretable if it directly answers the question “Which instances best exhibit the discrepancy between the samples?”

+

+ Where do the distributions differ. +

+

Outcomes

+
    +
  • Y. Karahan, B. Riaz, and A. J. Brockmeier “Kernel landmarks: An empirical statistical approach to detect covariate shift”, Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021) Paper, Poster, 5 min Presentation.
    • +
  • A. J. Brockmeier, C. C. Claros Olivares, M. S. Emigh, and L. G. Sanchez Giraldo “Identifying the instances associated with distribution shifts using the max-sliced Bures divergence”, Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021) Paper, Poster, 5 min Presentation.
    • +
+

Acknowledgements

+

These research efforts are sponsored by the Department of the Navy, Office of Naval Research under ONR award number N00014-21-1-2300. Gratefully acknowledge the support of Dr. Tory Cobb and ONR321. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Office of Naval Research.

+ +
+ + + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/research/eegs/index.html b/public/research/eegs/index.html new file mode 100644 index 0000000..824e558 --- /dev/null +++ b/public/research/eegs/index.html @@ -0,0 +1,245 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Learning Recurrent Waveforms Within EEGs

+

+ Austin J. Brockmeier and Jose C. Principe
+ IEEE Trans. on Biomedical Engineering, + Vol. 63, + No. 1, + pp. 43-54, + 2016 +

+ + + JOURNAL + + PDF + + CODE + +
+

Overview

+

When experts analyze EEGs they look for landmarks in the traces corresponding to established patterns such as oscillatory and phasic events of particular frequency or morphology. Long records motivate automated analysis techniques. Automation techniques often require design choices such as wavelet family or number of bandpass filters. To overcome this, we explore a modeling approach that automatically learns recurrent temporal waveforms within EEG traces. The estimation is based on a multiple-input, single-output linear model with sparsely excited inputs.

+

+ We assume that an EEG signal can be described by a convolutional sparse coding model. The EEG trace is approximated as an additive mixture of component signals each described by a convolution of a sparse source with a waveform. Although this is a linear model (multiple input and single output), to separate the components requires a nonlinear analysis. Additionally, when the waveforms are unknown, this blind source separation problem is even more challenging. +

+

We apply the approach to various data sets to better understand the specificity and consistency of the estimated waveforms. In particular, we cluster the waveforms found on different electrodes/channels and subjects.

+

+ Each cluster of waveforms is described by its centroid, the waveform nearest the centroid, the spectrum of the centroid, and the spatial distribution of the originating electrodes for the cluster waveforms. +

+ +
+ + + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/research/graphs/index.html b/public/research/graphs/index.html new file mode 100644 index 0000000..0c723a1 --- /dev/null +++ b/public/research/graphs/index.html @@ -0,0 +1,248 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Advancing Machine Learning for Neuroimaging Through Topology-Aware Signal Processing

+

+ University of Delaware Research Foundation–Strategic Initiative, Principal Investigator: Austin J. Brockmeier, Mentor: Gonzalo Arce, 11/2019–10/2021.
+ + + + + +

+ + + + + + + + +
+

The proposed work plan is to develop machine learning techniques to work directly with graph signal processing techniques in the context of neuroimaging. The goal is to leverage information in the form of the topology of the signal sensors or measurement locations to refine the neural signal representations in order to improve the statistical power of tests for distinguishing differences between conditions or stimuli. The project’s scope includes the formulation, mathematical and statistical analysis, and initial validation of the proposed methodology.

+

+ First 5 harmonics (Fourier transform) of the gray matter graph. +

+

Outcomes

+
    +
  • NER'21: Hassan Baker, A. J. Brockmeier “Local and Sparse Linear Causal Models for fMRI Resting-State Signals” Full-text (IEEE) | Poster PDF +*Machine Learning for Health (ML4H): Hassan Baker, A. J. Brockmeier “Exploring latent networks in resting-state fMRI using voxel-to-voxel causal modeling feature selection” Paper, Poster
    • +
+

Acknowledgements

+

Research was carried out with the support of the University of Delaware Research Foundation. Hassan Baker acknowledges support from the Gore Engineering +Graduate Fellowship in Spring 2021. This research was supported in part through the use of Data Science Institute (DSI) computational resources at the University of Delaware, specifically the DARWIN system. The DARWIN computing project at the University of Delaware is supported by the National Science Foundation under Grant No. OAC-1919839. The authors thank the support from the University of Delaware IT Research Computing Group.

+ +
+ + + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/research/index.html b/public/research/index.html new file mode 100644 index 0000000..42555cf --- /dev/null +++ b/public/research/index.html @@ -0,0 +1,357 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Research

+
+ +
+
News and Recent Research
+
+ + + +
+ + +

+ + + + + Learn More + +
+ +
+
Interpretable Maximal Discrepancies Metrics for Analyzing and Improving Generative Models
+
+ + + +
+ + +

Project funded by the Office of Naval Research

+ + + + + Learn More + +
+ +
+
Advancing Machine Learning for Neuroimaging Through Topology-Aware Signal Processing
+
+ + + +
+ + +

Project funding from a University of Delaware Research Foundation-Strategic Initiative grant

+ + + + + Learn More + +
+ +
+
Improving Reference Prioritisation With PICO Recognition
+
+ + + +
+ + +

Machine learning can assist with multiple tasks during systematic reviews to facilitate the rapid retrieval of relevant references during screening and to identify and extract information relevant to the study characteristics, which include the PICO elements of patient/population, intervention, comparator, and outcomes.

+ + + + + Learn More + +
+ +
+
Quantifying the Informativeness of Similarity Measurements
+
+ + + +
+ + +

In this paper, we describe an unsupervised measure for quantifying the 'informativeness' of correlation matrices formed from the pairwise similarities or relationships among data instances.

+ + + + + Learn More + +
+ +
+
Learning Recurrent Waveforms Within EEGs
+
+ + + +
+ + +

We explore a modeling approach that automatically learns recurrent temporal waveforms within EEG traces.

+ + + + + Learn More + +
+ +
+
Neural Decoding With Kernel-Based Metric Learning
+
+ + + +
+ + +

Machine learning (optimizing feature weightings and projections using kernel-based dependence) for enhancing neural data analysis, applied to a somatosensory neural decoding task.

+ + + + + Learn More + +
+ +
+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/research/index.xml b/public/research/index.xml new file mode 100644 index 0000000..0e9af9d --- /dev/null +++ b/public/research/index.xml @@ -0,0 +1,79 @@ + + + + Researches on CNIEL - University of Delaware + https://www.eecis.udel.edu/~ajbrock/research/ + Recent content in Researches on CNIEL - University of Delaware + Hugo -- gohugo.io + en-us + Tue, 24 Aug 2021 00:59:28 -0400 + + News and Recent Research + https://www.eecis.udel.edu/~ajbrock/research/news/ + Tue, 24 Aug 2021 00:59:28 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/news/ + Application Deadlines University of Delaware&rsquo;s Ph.D. program in Electrical and Computer Engineering (apply early; February 1 is the deadline to be considered for departmental funding.) University of Delaware&rsquo;s Undergraduate Summer Scholars (Please contact me ASAP as deadlines are usually Feb. or Mar.) University of Delaware&rsquo;s Latin and South America Summer Program for Visiting Researchers 2022 Summer Program Application Form Deadline: February 15th, 2022. December 2021 —Good luck to all the conference presenters this month: + + + + Interpretable maximal discrepancies metrics for analyzing and improving generative models + https://www.eecis.udel.edu/~ajbrock/research/discrepancies/ + Wed, 14 Apr 2021 00:59:28 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/discrepancies/ + Overview Divergence measures quantify the dissimilarity, including the distance, between distributions and are fundamental to hypothesis testing, information theory, and the estimation and criticism of statistical models. Recently, there has been renewed interest in divergences in the context of generative adversarial neural networks (GANs). While a multitude of divergences exist, they vary in their characteristics. Importantly, not all divergences are equally interpretable: a divergence between samples is considered interpretable if it directly answers the question “Which instances best exhibit the discrepancy between the samples? + + + + Advancing machine learning for neuroimaging through topology-aware signal processing + https://www.eecis.udel.edu/~ajbrock/research/graphs/ + Sat, 01 Feb 2020 00:59:28 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/graphs/ + The proposed work plan is to develop machine learning techniques to work directly with graph signal processing techniques in the context of neuroimaging. The goal is to leverage information in the form of the topology of the signal sensors or measurement locations to refine the neural signal representations in order to improve the statistical power of tests for distinguishing differences between conditions or stimuli. The project’s scope includes the formulation, mathematical and statistical analysis, and initial validation of the proposed methodology. + + + + Improving reference prioritisation with PICO recognition + https://www.eecis.udel.edu/~ajbrock/research/pico/ + Sat, 01 Jun 2019 00:59:28 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/pico/ + Overview +Machine learning can assist with multiple tasks during systematic reviews to facilitate the rapid retrieval of relevant references during screening and to identify and extract information relevant to the study characteristics, which include the PICO elements of patient/population, intervention, comparator, and outcomes. + +A publicly available corpus of PICO annotations on biomedical abstracts is used to train a named entity recognition model, which is implemented as a recurrent neural network. + + + + Quantifying the informativeness of similarity measurements + https://www.eecis.udel.edu/~ajbrock/research/informativeness/ + Fri, 17 Mar 2017 21:25:29 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/informativeness/ + Overview Choosing the particulars of a data representation is crucial for the successful application of machine learning techniques. In the unsupervised case, there is a lack of measures that can be used to compare different parameter choices that affect the representation. In this paper, we describe an unsupervised measure for quantifying the &lsquo;informativeness&rsquo; of correlation matrices formed from the pairwise similarities or relationships among data instances. +The measure quantifies the heterogeneity of the correlations and is defined as the distance between a correlation matrix and the nearest correlation matrix with constant off-diagonal entries. + + + + Learning Recurrent Waveforms within EEGs + https://www.eecis.udel.edu/~ajbrock/research/eegs/ + Thu, 28 Apr 2016 22:42:02 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/eegs/ + Overview When experts analyze EEGs they look for landmarks in the traces corresponding to established patterns such as oscillatory and phasic events of particular frequency or morphology. Long records motivate automated analysis techniques. Automation techniques often require design choices such as wavelet family or number of bandpass filters. To overcome this, we explore a modeling approach that automatically learns recurrent temporal waveforms within EEG traces. The estimation is based on a multiple-input, single-output linear model with sparsely excited inputs. + + + + Neural Decoding with Kernel-based Metric Learning + https://www.eecis.udel.edu/~ajbrock/research/neural_decoding/ + Mon, 28 Apr 2014 23:19:06 -0400 + + https://www.eecis.udel.edu/~ajbrock/research/neural_decoding/ + Overview Given a sample of data points, we often assume the points reside in some space in which we can measure distances between pairs of points. From these measurements we can understand which points are close to each other. Nearby points are often assumed to share characteristics. This assumption is the foundation of nearest-neighbor classification and regression as well as clustering. +The function that measures the distance between pairs of points in a particular space is called a distance metric. + + + + diff --git a/public/research/informativeness/index.html b/public/research/informativeness/index.html new file mode 100644 index 0000000..9075e1d --- /dev/null +++ b/public/research/informativeness/index.html @@ -0,0 +1,246 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Quantifying the Informativeness of Similarity Measurements

+

+ Austin J. Brockmeier, Tingting Mu, Sophia Ananiadou, and John Y. Goulermas
+ Journal of Machine Learning Research, + Vol. 18, + No. 18, + pp. 1-61, + 2017 +

+ + + JOURNAL + + PDF + + CODE + +
+

Overview

+

Choosing the particulars of a data representation is crucial for the successful application of machine learning techniques. In the unsupervised case, there is a lack of measures that can be used to compare different parameter choices that affect the representation. In this paper, we describe an unsupervised measure for quantifying the ‘informativeness’ of correlation matrices formed from the pairwise similarities or relationships among data instances.

+

The measure quantifies the heterogeneity of the correlations and is defined as the distance between a correlation matrix and the nearest correlation matrix with constant off-diagonal entries. While a homogenous correlation matrix indicates every instance is the same or equally dissimilar, informative correlation matrices are not uniform, some subsets of instances are more similar and themselves are dissimilar to other subsets. A set of distinct clusters is highly informative (Figure 1).

+

+ Figure 1: Informativeness versus von Neumann entropy for correlation matrices obtained from various configurations of four unit vectors. Both measures are minimal when the vectors are configured in a single cluster. Informativeness is higher for nontrivial clusterings, whereas entropy is maximized when the vectors are maximally separated. +

+

Informativeness can be used as an function to choose between representations or perform parameter selection (Figure 2) or dimensionality reduction. Using it, we designed a convex optimization algorithm for de-noising correlation matrices that clarifies their cluster structure.

+

+ Figure 2: Informativeness versus the von Neumann entropy of correlation matrices obtained from a Gaussian kernel applied with varying bandwidths to a sample with 2 clusters. +

+ +
+ + + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/research/neural_decoding/index.html b/public/research/neural_decoding/index.html new file mode 100644 index 0000000..4093baf --- /dev/null +++ b/public/research/neural_decoding/index.html @@ -0,0 +1,244 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Neural Decoding With Kernel-Based Metric Learning

+

+ Austin J. Brockmeier, John S. Choi, Evan G. Kriminger, Joseph T. Francis, and Jose C. Principe
+ Neural Computation, + Vol. 26, + No. 6, + pp. 1080-1107, + 2014 +

+ + + JOURNAL + + PDF + + CODE + +
+

Overview

+

Given a sample of data points, we often assume the points reside in some space in which we can measure distances between pairs of points. From these measurements we can understand which points are close to each other. Nearby points are often assumed to share characteristics. This assumption is the foundation of nearest-neighbor classification and regression as well as clustering.

+

The function that measures the distance between pairs of points in a particular space is called a distance metric.

+

+ In the original metric space, the white gold circle is closer to three black circles (two of which are equidistant from it). In a new metric space, changes in the vertical axes contribute more to the distance. Now, two gold circles are closer than the black circles. +

+

When studying the function of the nervous system, the choice of metric for the neural responses is a pivotal assumption. A well-suited distance metric enables neuroscientists to gauge the similarity of neural responses to various stimuli and assess the variability of responses to a repeated stimulus. In particular, neural spike train metrics have been used to quantify the information content carried by the timing of action potentials. While a number of metrics for individual neurons exist, a method to optimally combine single-neuron metrics into multi-neuron, or population-based, metrics is lacking. For time-locked neural responses, we pose the problem of the supervised optimization of multi-neuron metrics and other metrics (including those for local field potentials (LFPs), which are the electric potentials measured within brain tissue). The goal is to tune the metric to better predict the stimulus from the neural response. Predicting the stimulus from the response is known as neural decoding.

+

To guide the choice of metric learning, we propose to use centered alignment, a kernel-based dependence measure, to measure the correlation between the similarity in the data space (determined by the metric) and the similarity in the stimulus space. We provide mathematical formulation of the appropriate kernel and distance functions, as well as a MATLAB implementation of the batch and mini-batch optimization of the centered alignment metric learning (CAML).

+ +
+ + + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/research/news/index.html b/public/research/news/index.html new file mode 100644 index 0000000..91a3d28 --- /dev/null +++ b/public/research/news/index.html @@ -0,0 +1,305 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

News and Recent Research

+

+
+ + + + + +

+ + + + + + + + +
+
Application Deadlines
+
    +
  • University of Delaware’s Ph.D. program in Electrical and Computer Engineering (apply early; February 1 is the deadline to be considered for departmental funding.)
    • +
  • University of Delaware’s Undergraduate Summer Scholars (Please contact me ASAP as deadlines are usually Feb. or Mar.)
    • +
  • University of Delaware’s Latin and South America Summer Program for Visiting Researchers 2022 Summer Program Application Form Deadline: February 15th, 2022.
    • +
+
December 2021
+

—Good luck to all the conference presenters this month:

+
    +
  • ACNP: Kristina Holton, S. Y. Chan, A. J. Brockmeier, D. Öngür, and M-H. Hall “Exploring the influences of functional connectivity architecture on cortical thickness networks in patients with early psychosis”, 60th Annual Meeting of the American College of Neuropsychopharmacology, San Juan, Puerto Rico, 12/2021.
    • +
  • Machine Learning for Health (ML4H): Hassan Baker, A. J. Brockmeier “Exploring latent networks in resting-state fMRI using voxel-to-voxel causal modeling feature selection” Paper, Poster
    • +
  • Machine Learning for Health (ML4H): Kristina Holton, S. Y. Chan, A. J. Brockmeier, D. Öngür, and M-H. Hall “Exploring latent networks in resting-state fMRI using +voxel-to-voxel causal modeling feature selection”
    • +
  • DistShift @ NeurIPS: Yuksel Karahan, Bilal Riaz, and A. J. Brockmeier “Kernel landmarks: An empirical statistical approach to detect covariate shift”, Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021) Paper, Poster, 5 min Presentation.
    • +
  • DistShift @ NeurIPS: A. J. Brockmeier, Claudio Cesar Claros Olivares, M. S. Emigh, and L. G. Sanchez Giraldo “Identifying the instances associated with distribution shifts using the max-sliced Bures divergence”, Workshop on Distribution Shifts, 35th Conference on Neural Information Processing Systems (NeurIPS 2021) Paper, Poster, 5 min Presentation.
    • +
+

—Carlos presents his proposal “Learning Representative Waveforms to Analyze, Summarize, and Compare Long-term Neural Recordings”

+
November 2021
+

—Kris wins best (joint) poster award at DSI’s 2021 Delaware Data Science Symposium with her work “Exploring the influences of functional connectivity architecture on cortical thickness networks in patients with early psychosis” Poster, Lightning presentation

+

—Carlos presented his work at EMBC'21, “Shift-invariant waveform learning on epileptic ECoG” PDF

+
October 2021
+

—Kristina (Kris) Holton, Certificate Program for Bioinformatics and Data Science, abstract and extended abstract accepted to Annual Meeting of the American College of Neuropsychopharmacology (ACNP) and Machine Learning for Health (ML4H).

+

—Hassan’s extended abstract also accepted to Machine Learning for Health (ML4H).

+

—César’s paper and Yüksel and Bilal’s paper accepted to DistShift at NeurIPS 2021.

+
September 2021
+

—Hassan passes the Ph.D. qualifying examination.

+

—Dr. Brockmeier will serve as Co-Investigator on a NIH-funded R21 project “Integrative Data Analysis to Identify Persistent Post-Concussion Deficits and Subsequent Musculoskeletal Injury Risk” led by Dr. Thomas Buckley

+

—NSF funds a “Harnessing the Data Revolution” traineeship on “Computing and Data Science Training for Materials Innovation, Discovery, Analytics” (MIDAS) led by Dr. Arthi Jayaraman. (Dr. Brockmeier will serve as senior personnel.)

+
    +
  • 1st year Ph.D. students interested in using AI, machine learning, and data science to solve problems in polymeric material design
    • +
  • 2nd year Ph.D. students will have a pilot offering of the problem-focused team oriented ‘hackathon’ course
    • +
+

— NSF funds a “Harnessing the Data Revolution” undergraduate research and training “Delaware and Mid-Atlantic Data Science Corps” NSF 2123264 led by Dr. Federica Bianco et al. (Dr. Brockmeier will serve as senior personnel.)

+
August 2021
+

—Summer Scholars (Evan Curtin and Justin Labombard) present at the Undergraduate Research and Service Scholar Celebratory Symposium

+
    +
  • Evan Curtin, “Non-Negative Matrix Factorization as Dictionary Learning for Audio Separation” Poster PDF
    • +
  • Justin Labombard, “Dictionary Learning on Eplieptic ECoG Waveforms” Poster PDF +—Kick off of new NSF-funded project led by Dr. Federica Bianco on AI for astronomy: ``Detecting and studying light echoes in the era of Rubin and artificial intelligence'' NSF 2108841
    • +
+
July 2021
+

—Andres Nicolas Lopez, MSc. (National University of Colombia), joins us as a visiting research working on modeling the uncertainty and errors of machine learning models University of Delaware’s Latin and South America Summer Program +—Carlos’s second paper accepted to EMBC'21

+
    +
  • “Shift-invariant waveform learning on epileptic ECoG” Pre-print PDF
    • +
+
July 2021
+

—Evan, Justin, and Thomas start the (undergraduate) Summer Scholars Program

+
May 2021
+

—Carlos and Hassan presented their work as posters at NER'21

+
    +
  • Hassan presented “Local and Sparse Linear Causal Models for fMRI Resting-State Signals” Full-text (IEEE) | Poster PDF
    • +
  • Carlos presented “Searching for Waveforms on Spatially-Filtered Epileptic ECoG” Full-text (IEEE) | PDF | Poster PDF
    • +
  • We made the top-5 finalist for the Neural Engineering Community Award
    • +
+
April 2021
+

—Kick off of new ONR-funded project on interpretable divergences

+ +
January 2021
+

— Hassan awarded Gore Engineering Graduate Fellowship for Spring 2021.

+

—Hau Phan joins the lab to work on a collaborative ONR-funded project for bioinspired processing of chemical signals.

+
September 2020
+

Bilal passes the Ph.D. qualifying examination.

+
August 2020
+

César completes his Master’s thesis “Synergistic Human-Machine Prediction: Active Error Analysis and Mitigation with Gaussian Process Regression”.

+
February 2020
+

Hassan joins the lab to work on a neuroimaging and graph signal processing project.

+ +
+ + + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/research/pico/index.html b/public/research/pico/index.html new file mode 100644 index 0000000..03f4229 --- /dev/null +++ b/public/research/pico/index.html @@ -0,0 +1,254 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Improving Reference Prioritisation With PICO Recognition

+

+ Austin J. Brockmeier, Meizhi Ju, Piotr Przybyła, Sophia Ananiadou
+ BMC Medical Informatics and Decision Making, + Vol. 19, + No. 1, + pp. 1–14, + 2019 +

+ + + JOURNAL + + PDF + + + +
+

Overview

+

+ Systematic reviews of biomedical research focus on specific questions framed in terms of the patient population, intervention, control, and outcomes measured (PICO). These terms are used to query literature databases. The resulting hits (which could be hundreds or thousands of articles) need to be sorted into relevant and irrelevant groups. +

+

Machine learning can assist with multiple tasks during systematic reviews to facilitate the rapid retrieval of relevant references during screening and to identify and extract information relevant to the study characteristics, which include the PICO elements of patient/population, intervention, comparator, and outcomes.

+

+ Example of automatically extracted participants, intervention, and outmodes (PIO). Techniques similar to named entity recognition are used for identifying and categorising fragments of text. +

+

A publicly available corpus of PICO annotations on biomedical abstracts is used to train a named entity recognition model, which is implemented as a recurrent neural network. This model is then applied to a separate collection of abstracts for references from systematic reviews within biomedical and health domains. The occurrences of words tagged in the context of specific PICO contexts are used as additional features for a relevancy classification model. Simulations of the machine learning-assisted screening are used to evaluate the work saved by the relevancy model with and without the PICO features.

+

+ PICO recognition and abstract screening process. In the first phase, the PICO recognition model is trained to predict the PICO mention spans +on a human annotated corpus of abstracts. In the second phase, a collection of abstracts is processed by the PICO recognition model and the +results along with the original abstract are used to create a vector representation of each abstract. In the final phase, a user labels abstracts as being +included (relevant) or excluded, these decisions are used to train a machine learning (ML) model that uses the vector representation. The ML model +is applied to the remaining unlabelled abstracts, which are then sorted by their predicted relevancy, the user sees the top ranked abstracts, labels +them, and this process repeats +

+

Chi-squared and statistical significance of positive predicted values are used to identify words that are more indicative of relevancy within PICO contexts. Inclusion of PICO features improves the performance metric on 15 of the 20 collections, with substantial gains on certain systematic reviews. Examples of words whose PICO context are more precise can explain this increase. Words within PICO tagged segments in abstracts are predictive features for determining inclusion. Combining PICO annotation model into the relevancy classification pipeline is a promising approach.

+ +
+ + + + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/singles/admissions/index.html b/public/singles/admissions/index.html new file mode 100644 index 0000000..d2f2afc --- /dev/null +++ b/public/singles/admissions/index.html @@ -0,0 +1,236 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Admissions

+

I will be accepting graduate students again (after the pandemic) starting Spring 2022 or Fall 2022 and continuing through subsequent semesters. In exceptional situations, support before the fall term.

+

Due to the fact that I may miss your email, please apply to the Department of Electrical and Computer Engineering (ECE) and designate me as a potential advisor. I am also able to advise students in the Department of Computer and Information Sciences (CIS). Both departments are in the College of Engineering at the University of Delaware. Starting Fall 2022, we are also specifically looking for students interested in pursuing a Ph.D. focused on data science and computing applied to polymer design and innovation. Find out more

+

Also starting in Fall 2022, UD is recruiting for its interdisciplinary neuroscience graduate (ING) program that spans many colleges including College of Engineering. Applicants with backgrounds in computational and analytic fields who are interested in advancing data science tools for understanding the brain are encouraged to apply.

+

In CIS, the admission review is performed by a committee. Applicants are encouraged to have a focused statement describing their research pursuits and their skills and experiences.

+

In ECE, the admission review also involves the associate chair of graduate studies. The process has many stages:

+

Applications received by Feb. 1 or earlier will receive primary consideration for funding. The research statement is important during the evaluation.

+

Applications are reviewed along with those for other suitable advisors.

+
    +
  • (January) Exceptional candidates may be contacted.
    • +
  • (Early February) An assessment of the applications is made and a subset of the qualified applicants may be short-listed.
    • +
  • (Feb–March) I typically schedule and hold short interviews with short-listed candidates on video teleconferencing along with current members of my group. I encourage applicants to contact and discuss the research and lab culture with current students.
    • +
  • (March) Depending on funding, I along with the associate chair of graduate studies make the final determination to offer funding for a certain duration.
    • +
  • (April 15th) Applicants should review the offer and accept it, or decline it if they choose to take another option. The deadline for accepting an offer of financial support from a graduate school in the US is April 15th. (PDF)
    • +
+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/singles/contact/index.html b/public/singles/contact/index.html new file mode 100644 index 0000000..9f03cd0 --- /dev/null +++ b/public/singles/contact/index.html @@ -0,0 +1,236 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Contact

+
+
+ +
+
+
Location
+

+ Evans Hall, Room 306
+ 139 The Green, Newark, Delaware 19716 +

+ + +
Email
+ + ajbrock@udel.edu + + + +
For admissions
+ + See here + +
+
+ + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/singles/education/index.html b/public/singles/education/index.html new file mode 100644 index 0000000..255b74e --- /dev/null +++ b/public/singles/education/index.html @@ -0,0 +1,304 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

Education

+

Fall semesters: second-year graduate-level course in machine learning.

+

Spring semesters: 300-level undergraduate course in signal processing and linear system analysis

+

Summer 2021: Independent study on adaptive filters

+

Spring 2021: Graduate Data Science Community Hour during Spring 2020

+

Graduate Course

+

Large-Scale Machine Learning ELEG/FSAN 817

+

Course Description: Introduction to the analysis and processing of massive high-dimensional data. Massive data sets generally involve growth not only in the number of individuals represented but also in the number of descriptive parameters of the individuals, leading to exponential growth in the number of hypothesis considered. Approaches to address these problems exploit concepts from statistics and machine learning.

+

Offered: Fall 2019, Fall 2020, Fall 2021

+

The course serves as a capstone course in machine learning for graduate students, with the goal of readying students for machine learning and data science research. Computational and statistical scaling of machine learning is discussed from both theoretical and practical perspectives. The formative assessments provide exercise in key aspects of machine learning research: problem formulation, critical reading of literature, algorithm implementation, abstract writing, experimental design, and peer review. Summative assessment is based on a mid-term exam and final project.

+

2020 online synchronous course enrollment: There were 17 students enrolled: ECE (9 Ph.D. / 2 M.S.), FSAN (3 Ph.D.), Computer Science (1 Ph.D.), and Physics and Astronomy (1 Ph.D.). One additional ECE Ph.D. student audited.

+

2019 course enrollment: There were 13 students enrolled: ECE (3 Ph.D. / 2 M.S. / 1 B.S./M.S.), Computer Science (5 Ph.D.), Economics (1 Ph.D.), Civil & Environmental Eng. (1 Ph.D., auditing).

+

Anonymous feedback from 2020 course evaluations:

+
    +
  • “The topics covered in the course are interesting and relevant. The instructor exhibits a deep understanding of the subject and has the capacity to motivate interest in the students.”
    • +
  • “The instructor did a great job structuring the course and selecting the key concepts and methods among the vast amount of topics in large-scale machine learning. He is very knowledgeable and demonstrates an honest interest in our learning experience. The evaluation tools used (the midterm, the pop quizzes, the homeworks, and the final project) are well designed to accomplish our learning goals. I particularly liked that he made emphasis throughout the course in that the evaluation tools were more intended as learning tools, and offer us options to ease the natural pressure that the grades associated with those tools usually have.”
    • +
  • “I like the discussions and that despite being online, the professor was capable of keeping me engaged even when the topics were not particularly easy to explain. I appreciate the use of a whiteboard instead of the slides, it gives flexibility to the class and it helps me to understand.”
    • +
  • “He always wants us understand the topics. He supports lecture[…] interactions. He asks during lectures to make sure we are on the right track. He tried to link topics that are correlated and related to each other. He links the topics we are studying to a real world examples.”
    • +
  • “I can see Prof. Brockmeier really wants students to learn, no matter whether a student is already a machine learning research[er] in a specific area or is still exploring the subject.”
    • +
  • “Dr. Brockmeier is a great professor. He was well prepared for the classes and was capable of address almost every question that pops up during the course and always made a good connection between theory and Matlab examples.”
    • +
  • “There will be timely comments and replies to every student’s questions and papers”
    • +
  • “He was open to listening to the questions of all the students that wanted to ask and address those questions in a respectful way.”
    • +
  • “He was always polite and gave everyone the opportunity to express their opinions and ideas regarding the topics that were explained.”
    • +
  • “I rank him in the top 1 percentile for research ability, approachability and overall being a good person. I would take his class again and recommend it people.”
    • +
+

Anonymous feedback from 2019 course evaluations:

+
    +
  • “The course really dives deep into the mathematics and theory of the machine learning techniques. The organization of the course was nice.”
    • +
  • “Course material was well ordered, well motivated, and useful.”
    • +
  • “The breadth of topics studied in this class provided me with a better understanding of machine learning in general.”
    • +
  • “Professor Brockmeier […] was able to connect abstract concepts from the class to multiple applications to motivate understanding”
    • +
  • “This was not only the best course I took as a graduate student, but moreover, the most useful in the pursuit of becoming a machine learning expert.”
    • +
  • “He was enthusiastic about delivering the contents of the course.”
    • +
+

Undergraduate Courses

+

Signals and Systems ELEG 305

+

This course examines continuous and discrete-time signals and systems at the introductory +level.

+

Course description: Introduction to signals and systems, with an emphasis on time and frequency characterization of linear, time-invariant systems. Covers discrete and continuous time systems, sampling, and Fourier, Laplace, and Z transforms.

+

Offered: Spring 2020, +Spring 2021, Spring 2022

+

Core requirement in the following programs:

+
    +
  • Electrical Engineering
    • +
  • Computer Engineering +and computer engineering majors.
    • +
  • GIScience and Environmental Data Analytics
    • +
+

Technical elective:

+
    +
  • Chemical Engineering, Materials Science and Engineering, Mechanical Engineering
    • +
+

Prerequisite for courses:

+
    +
  • ELEG 306 - Digital Signal Processing
    • +
  • ELEG 403 - Communication Systems Engineering
    • +
  • ELEG 404 - Digital Imaging and Photography
    • +
  • ELEG 418 - Digital Control Systems
    • +
  • ELEG 492 - Radar Systems and Technology
    • +
  • BMEG 441 - Biomechatronics†
    • +
  • BMEG 330 - Biomedical Instrumentation†
    +† As an alternative for BMEG 230
    • +
+

Honors Discussion Section: +Students registered in Honors for this class will be involved in an additional discussion section, which introduces content beyond the general class: specifically, non-linear, adaptive, and random systems. They will also be required to complete a semester long project that reflects an application of such systems.

+

Unsolicited Feedback from 2021 Honors section:

+
    +
  • “I really enjoyed looking more into something I’m very interested in. It was my favorite Zoom meeting between all of my classes this semester. Connecting with what I focused on, […] Doing Honors research got me to realize that dream.”
    • +
+

Unsolicited Feedback from 2020 section:

+
    +
  • “I want to applaud your tenacity and optimism throughout the term. Despite this class being the most challenging course I’ve ever taken, I have enjoyed being your student.
    • +
+

Independent Study ELEG 466

+

“Adaptive Filtering as an Introduction to Machine Learning”

+

Offered: Summer 2021

+

Course description: Introduction to adaptive filter theory and application. Students will create a technical report that details the theory, implementation, and computational experiments with the application of adaptive filtering. Topics to be covered include review of delay lines, z-transform, multivariate signal processing, Gaussian white noise, correlation matrices, whitening transforms, least squares, convex optimization, Wiener filter, stochastic gradient descent, least mean squares (LMS), NLMS, common spatial patterns, and various types of recurrent memory. A comprehensive technical report with equations and figures will be typeset in Latex. The report will detail the problem formulations, equations, and vector graphics of block diagrams, plots of results. Software implementations will be made in MATLAB/Octave, Julia, python/Numpy, or python/Pytorch. Synthetic experiments will be conducted for noise cancellation, channel estimation, and machine learning.

+

Learning outcomes: +While students typically learn theory and fundamentals of signal processing and filtering in ELEG305 “Signals and Systems” and probability theory and statistics related to signals in ELEG310, due to time constraints and curriculum design, students rarely have the opportunity to learn and apply adaptive filter theory that combines filtering, linear algebra, optimization, and statistics together. This combination is the foundation of machine learning theory and algorithms. This independent study will attempt to fill the gap in the current curriculum areas in theory and practice. A combination of synthetic data and real-world projects will be explored by students to master these concepts. Students will practice their understanding by describing the formulations and results (equation and paper typesetting, block diagrams, data visualization and presentation) in their own words as well as implementing and testing software for adaptive filtering and machine learning. In summary the students will learn the foundations and techniques of machine learning: linear and non-linear filtering, supervised learning, least squares, optimization, stochastic gradient descent, batch normalization, and experimental design (training, validation, and testing).

+

Experience: +The course will solidify knowledge learned in other courses and provide the student with a diverse and broad understanding of adaptive filtering, and an introduction to machine learning. The course project will require tools learned in other academic classes (signal and systems, linear algebra, calculus, probability, and statistics) and will help the student in future classes (communication systems, machine learning, control theory) as well as research.

+

Expectations: +Students will be required to complete a technical report which will be a culmination of most of the topics covered in class. Weekly revised versions of the technical report must be submitted to document learning progress until the end of the course. Each of these weekly submissions, will require an executive summary that overviews the specific learning activities, progress in understanding and experiments, obstacles, and interesting insights or useful skills that have been learned. Students can work in groups and collaborate in a civil and equal manner.

+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/singles/index.html b/public/singles/index.html new file mode 100644 index 0000000..9307643 --- /dev/null +++ b/public/singles/index.html @@ -0,0 +1,182 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

This is Singles, list page

+ + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/singles/index.xml b/public/singles/index.xml new file mode 100644 index 0000000..164419b --- /dev/null +++ b/public/singles/index.xml @@ -0,0 +1,48 @@ + + + + Singles on CNIEL - University of Delaware + https://www.eecis.udel.edu/~ajbrock/singles/ + Recent content in Singles on CNIEL - University of Delaware + Hugo -- gohugo.io + en-us + Tue, 01 Jun 2021 02:38:52 -0400 + + Education + https://www.eecis.udel.edu/~ajbrock/singles/education/ + Tue, 01 Jun 2021 02:38:52 -0400 + + https://www.eecis.udel.edu/~ajbrock/singles/education/ + Fall semesters: second-year graduate-level course in machine learning. Spring semesters: 300-level undergraduate course in signal processing and linear system analysis Summer 2021: Independent study on adaptive filters Spring 2021: Graduate Data Science Community Hour during Spring 2020 Graduate Course Large-Scale Machine Learning ELEG/FSAN 817 Course Description: Introduction to the analysis and processing of massive high-dimensional data. Massive data sets generally involve growth not only in the number of individuals represented but also in the number of descriptive parameters of the individuals, leading to exponential growth in the number of hypothesis considered. + + + + K–12 Outreach and Undergraduate Research + https://www.eecis.udel.edu/~ajbrock/singles/outreach/ + Thu, 29 Apr 2021 00:49:51 -0400 + + https://www.eecis.udel.edu/~ajbrock/singles/outreach/ + In 2019, we joined UD&rsquo;s Women in Engineering outreach program at a local New Castle Middle School, Serviam Girls Academy , whose mission is to take girls from low-income houses and give them the education they need to succeed in high school and beyond. Working with Trevor Foresta, ECE undergraduate student and Julia Schneider, a post-doctoral researcher from the Department of Linguistics and Cognitive Sciences, we developed a module showing ECE’s critical role in neuroimaging technology and analysis for their enrichment program during their enrichment hour (4-5 pm on Wednesday). + + + + Admissions + https://www.eecis.udel.edu/~ajbrock/singles/admissions/ + Wed, 28 Apr 2021 12:49:16 -0400 + + https://www.eecis.udel.edu/~ajbrock/singles/admissions/ + I will be accepting graduate students again (after the pandemic) starting Spring 2022 or Fall 2022 and continuing through subsequent semesters. In exceptional situations, support before the fall term. +Due to the fact that I may miss your email, please apply to the Department of Electrical and Computer Engineering (ECE) and designate me as a potential advisor. I am also able to advise students in the Department of Computer and Information Sciences (CIS). + + + + Contact + https://www.eecis.udel.edu/~ajbrock/singles/contact/ + Wed, 28 Apr 2021 12:49:01 -0400 + + https://www.eecis.udel.edu/~ajbrock/singles/contact/ + + + + + diff --git a/public/singles/outreach/index.html b/public/singles/outreach/index.html new file mode 100644 index 0000000..15937c4 --- /dev/null +++ b/public/singles/outreach/index.html @@ -0,0 +1,252 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

K–12 Outreach and Undergraduate Research

+

+ Outreach Image +

+

In 2019, we joined UD’s Women in Engineering outreach program at a local New Castle Middle School, Serviam Girls Academy , whose mission is to take girls from low-income houses and give them the education they need to succeed in high school and beyond. Working with Trevor Foresta, ECE undergraduate student and Julia Schneider, a post-doctoral researcher from the Department of Linguistics and Cognitive Sciences, we developed a module showing ECE’s critical role in neuroimaging technology and analysis for their enrichment program during their enrichment hour (4-5 pm on Wednesday). Julia provided slides and enthusiasm on current and past sensors and systems to measure the brains operation. Trevor assembled and tested the OpenBCI headset while still a first-year engineering student.

+

+ Outreach Image +
+
+ Outreach Image +
+
+ Outreach Image +

+

In 2020, Trevor and I (along with other volunteers) hosted a lunchtime activity booth at the “Engineering Your Tomorrow” event held at Delaware Tech Community College in Georgetown, DE. The Sussex County STEM Alliance organized this outreach event with the purpose “to expose, mentor and teach our County’s diverse young girls the basic fundamentals of science and engineering through high interest STEM activities, while showing them… “STEM IS FUN”. After helping with the entomology experiments in the morning, I set up the booth that showcased ECE’s critical role in biomedical engineering with a live demonstration and interactive handouts. Thanks to UD’s Erin Sparks for inviting us to this event.

+

+ Outreach Image +
+
+ Outreach Image +
+
+ Outreach Image +

+

+
+

Neural Engineering Community Award

+
+
+ +Our initial outreach efforts were recognized by being one of five short-listed groups for the Neural Engineering Community Award at the 2021 Neural Engineering Workshop (NER'21)*, a biennial conference hosted by the IEEE Engineering in Medicine & Biology Society. The award was for outstanding contributions to the mission of inclusion and diversity. Congratulations to the winners! We will increase our impact and efforts in 2021–2022.

+ + + + + + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/sitemap.xml b/public/sitemap.xml new file mode 100644 index 0000000..cdd9675 --- /dev/null +++ b/public/sitemap.xml @@ -0,0 +1,143 @@ + + + + + https://www.eecis.udel.edu/~ajbrock/ + 2021-08-24T00:59:28-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/research/news/ + 2021-08-24T00:59:28-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/research/ + 2021-08-24T00:59:28-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/ + 2021-06-20T17:36:38-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/nicolaslopez/ + 2021-06-20T17:36:38-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/justinlabombard/ + 2021-06-15T17:00:22-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/evancurtin/ + 2021-06-01T21:00:00-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/singles/education/ + 2021-06-01T02:38:52-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/singles/ + 2021-06-01T02:38:52-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/singles/outreach/ + 2021-04-29T00:49:51-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/singles/admissions/ + 2021-04-28T12:49:16-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/singles/contact/ + 2021-04-28T12:49:01-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/research/discrepancies/ + 2021-04-14T00:59:28-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/hauphan/ + 2021-01-01T17:00:52-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/hassanbaker/ + 2020-02-01T17:00:38-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/research/graphs/ + 2020-02-01T00:59:28-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/bilalriaz/ + 2019-08-28T17:00:22-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/yukselkarahan/ + 2019-08-28T16:59:19-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/edwinsalcedo/ + 2019-07-01T17:36:38-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/research/pico/ + 2019-06-01T00:59:28-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/cesarolivares/ + 2019-05-01T17:00:02-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/carloscardenas/ + 2019-01-01T13:12:57-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/members/austinbrockmeier/ + 2018-04-28T17:07:03-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/research/informativeness/ + 2017-03-17T21:25:29-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/research/eegs/ + 2016-04-28T22:42:02-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/research/neural_decoding/ + 2014-04-28T23:19:06-04:00 + + + + https://www.eecis.udel.edu/~ajbrock/categories/ + + + + https://www.eecis.udel.edu/~ajbrock/tags/ + + + diff --git a/public/tags/index.html b/public/tags/index.html new file mode 100644 index 0000000..d90b1c1 --- /dev/null +++ b/public/tags/index.html @@ -0,0 +1,182 @@ + + + + + + + + + + + + + + + + + + + + + +
+
+ + + +

This is Tags, list page

+ + +
+
+ + + +
+ + +
+ Made with Hugo | Designed by Zihan Wu | Copyright © 2021 Austin J. Brockmeier +
+ + + + + + diff --git a/public/tags/index.xml b/public/tags/index.xml new file mode 100644 index 0000000..de10ed6 --- /dev/null +++ b/public/tags/index.xml @@ -0,0 +1,10 @@ + + + + Tags on CNIEL - University of Delaware + https://www.eecis.udel.edu/~ajbrock/tags/ + Recent content in Tags on CNIEL - University of Delaware + Hugo -- gohugo.io + en-us + + diff --git a/resources/.DS_Store b/resources/.DS_Store new file mode 100644 index 0000000..b14aade Binary files /dev/null and b/resources/.DS_Store differ diff --git a/resources/_gen/.DS_Store b/resources/_gen/.DS_Store new file mode 100644 index 0000000..b697cc1 Binary files /dev/null and b/resources/_gen/.DS_Store differ diff --git a/static/.DS_Store b/static/.DS_Store new file mode 100644 index 0000000..d0636bd Binary files /dev/null and b/static/.DS_Store differ diff --git a/static/code/CAML.m b/static/code/CAML.m old mode 100644 new mode 100755 diff --git a/static/code/CAML_approx.m b/static/code/CAML_approx.m old mode 100644 new mode 100755 diff --git a/static/code/cmd_simple_test.m b/static/code/cmd_simple_test.m old mode 100644 new mode 100755 diff --git a/static/code/eegs.zip b/static/code/eegs.zip old mode 100644 new mode 100755 diff --git a/static/code/informativeness.zip b/static/code/informativeness.zip old mode 100644 new mode 100755 diff --git a/static/code/learning_recurrent_waveforms_v1.zip b/static/code/learning_recurrent_waveforms_v1.zip old mode 100644 new mode 100755 diff --git a/static/code/neural_decoding.zip b/static/code/neural_decoding.zip old mode 100644 new mode 100755 diff --git a/static/code/shiftInvariantKmeans.m b/static/code/shiftInvariantKmeans.m old mode 100644 new mode 100755 diff --git a/static/css/style.css b/static/css/style.css old mode 100644 new mode 100755 index e981bc1..5e4f83d --- a/static/css/style.css +++ b/static/css/style.css @@ -202,4 +202,4 @@ footer{ text-align: center; margin: 0; } -} +} \ No newline at end of file diff --git a/static/images/.DS_Store b/static/images/.DS_Store new file mode 100644 index 0000000..9873ea4 Binary files /dev/null and b/static/images/.DS_Store differ diff --git a/static/images/cover-slide.png b/static/images/cover-slide.png new file mode 100644 index 0000000..9e4df2f Binary files /dev/null and b/static/images/cover-slide.png differ diff --git a/static/images/evanshall.png b/static/images/evanshall.png old mode 100644 new mode 100755 diff --git a/static/images/logo-black-fill.svg b/static/images/logo-black-fill.svg old mode 100644 new mode 100755 diff --git a/static/images/logo-black-outline.svg b/static/images/logo-black-outline.svg old mode 100644 new mode 100755 diff --git a/static/images/logo-white-fill.svg b/static/images/logo-white-fill.svg old mode 100644 new mode 100755 diff --git a/static/images/logo-white-outline.svg b/static/images/logo-white-outline.svg old mode 100644 new mode 100755 diff --git a/static/images/members/.DS_Store b/static/images/members/.DS_Store new file mode 100644 index 0000000..0db98f5 Binary files /dev/null and b/static/images/members/.DS_Store differ diff --git a/static/images/members/AustinBrockmeier.jpg b/static/images/members/AustinBrockmeier.jpg old mode 100644 new mode 100755 diff --git a/static/images/members/Carlos.jpeg b/static/images/members/Carlos.jpeg new file mode 100644 index 0000000..eed8596 Binary files /dev/null and b/static/images/members/Carlos.jpeg differ diff --git a/static/images/members/Cesar.jpg b/static/images/members/Cesar.jpg new file mode 100644 index 0000000..427530a Binary files /dev/null and b/static/images/members/Cesar.jpg differ diff --git a/static/images/members/Edwin.png b/static/images/members/Edwin.png new file mode 100644 index 0000000..f86302f Binary files /dev/null and b/static/images/members/Edwin.png differ diff --git a/static/images/members/EvanCurtin.png b/static/images/members/EvanCurtin.png new file mode 100644 index 0000000..fef8b33 Binary files /dev/null and b/static/images/members/EvanCurtin.png differ diff --git a/static/images/members/Hassan.png b/static/images/members/Hassan.png new file mode 100644 index 0000000..842c62c Binary files /dev/null and b/static/images/members/Hassan.png differ diff --git a/static/images/members/Hau.gif b/static/images/members/Hau.gif new file mode 100644 index 0000000..b74f8f1 Binary files /dev/null and b/static/images/members/Hau.gif differ diff --git a/static/images/members/Justin.png b/static/images/members/Justin.png new file mode 100644 index 0000000..948d053 Binary files /dev/null and b/static/images/members/Justin.png differ diff --git a/static/images/members/Nicolas.jpg b/static/images/members/Nicolas.jpg new file mode 100644 index 0000000..a9f32ea Binary files /dev/null and b/static/images/members/Nicolas.jpg differ diff --git a/static/images/members/yuksel.png b/static/images/members/yuksel.png new file mode 100644 index 0000000..d3709d3 Binary files /dev/null and b/static/images/members/yuksel.png differ diff --git a/static/images/outreach/image0.png b/static/images/outreach/image0.png new file mode 100644 index 0000000..4123bb0 Binary files /dev/null and b/static/images/outreach/image0.png differ diff --git a/static/images/outreach/image1.jpg b/static/images/outreach/image1.jpg old mode 100644 new mode 100755 diff --git a/static/images/outreach/image3.jpg b/static/images/outreach/image3.jpg old mode 100644 new mode 100755 diff --git a/static/images/outreach/image4.jpg b/static/images/outreach/image4.jpg old mode 100644 new mode 100755 diff --git a/static/images/outreach/image5.jpg b/static/images/outreach/image5.jpg old mode 100644 new mode 100755 diff --git a/static/images/outreach/image6.jpg b/static/images/outreach/image6.jpg old mode 100644 new mode 100755 diff --git a/static/images/outreach/image7.jpg b/static/images/outreach/image7.jpg old mode 100644 new mode 100755 diff --git a/static/images/outreach/image8.jpg b/static/images/outreach/image8.jpg old mode 100644 new mode 100755 diff --git a/static/images/profilePlaceholder.svg b/static/images/profilePlaceholder.svg old mode 100644 new mode 100755 diff --git a/static/images/research/.DS_Store b/static/images/research/.DS_Store new file mode 100644 index 0000000..5008ddf Binary files /dev/null and b/static/images/research/.DS_Store differ diff --git a/static/images/research/brainGraph1.png b/static/images/research/brainGraph1.png new file mode 100644 index 0000000..4d7df23 Binary files /dev/null and b/static/images/research/brainGraph1.png differ diff --git a/static/images/research/brainGraph2.png b/static/images/research/brainGraph2.png new file mode 100644 index 0000000..15b1357 Binary files /dev/null and b/static/images/research/brainGraph2.png differ diff --git a/static/images/research/discrepancies.png b/static/images/research/discrepancies.png new file mode 100644 index 0000000..3f99056 Binary files /dev/null and b/static/images/research/discrepancies.png differ diff --git a/static/images/research/discrepancies2.png b/static/images/research/discrepancies2.png new file mode 100644 index 0000000..5b62891 Binary files /dev/null and b/static/images/research/discrepancies2.png differ diff --git a/static/images/research/eegs1.png b/static/images/research/eegs1.png old mode 100644 new mode 100755 diff --git a/static/images/research/eegs2.png b/static/images/research/eegs2.png new file mode 100644 index 0000000..7d3d3b5 Binary files /dev/null and b/static/images/research/eegs2.png differ diff --git a/static/images/research/informativeness1.png b/static/images/research/informativeness1.png old mode 100644 new mode 100755 diff --git a/static/images/research/informativeness2.png b/static/images/research/informativeness2.png old mode 100644 new mode 100755 diff --git a/static/images/research/metricLearning.png b/static/images/research/metricLearning.png new file mode 100644 index 0000000..0d366ea Binary files /dev/null and b/static/images/research/metricLearning.png differ diff --git a/static/images/research/metricLearning_thumb.png b/static/images/research/metricLearning_thumb.png new file mode 100644 index 0000000..30d9df3 Binary files /dev/null and b/static/images/research/metricLearning_thumb.png differ diff --git a/static/images/research/neuralDecoding1.png b/static/images/research/neuralDecoding1.png old mode 100644 new mode 100755 diff --git a/static/images/research/pico1.png b/static/images/research/pico1.png new file mode 100644 index 0000000..98c0065 Binary files /dev/null and b/static/images/research/pico1.png differ diff --git a/static/images/research/pico2.png b/static/images/research/pico2.png new file mode 100644 index 0000000..3f0e015 Binary files /dev/null and b/static/images/research/pico2.png differ diff --git a/static/images/research/pico3.png b/static/images/research/pico3.png new file mode 100644 index 0000000..99b2d81 Binary files /dev/null and b/static/images/research/pico3.png differ diff --git a/static/images/research/pico4.png b/static/images/research/pico4.png new file mode 100644 index 0000000..d7b689a Binary files /dev/null and b/static/images/research/pico4.png differ diff --git a/static/other/ajbrockmeier_CV_12_2021.pdf b/static/other/ajbrockmeier_CV_12_2021.pdf new file mode 100644 index 0000000..21d44b4 Binary files /dev/null and b/static/other/ajbrockmeier_CV_12_2021.pdf differ diff --git a/static/other/ajbrockmeier_CV_8_2021.pdf b/static/other/ajbrockmeier_CV_8_2021.pdf new file mode 100644 index 0000000..2e8709d Binary files /dev/null and b/static/other/ajbrockmeier_CV_8_2021.pdf differ diff --git a/static/other/baker_poster_ML4H2021.pdf b/static/other/baker_poster_ML4H2021.pdf new file mode 100644 index 0000000..e887e81 Binary files /dev/null and b/static/other/baker_poster_ML4H2021.pdf differ diff --git a/static/other/baker_poster_NER2021.pdf b/static/other/baker_poster_NER2021.pdf new file mode 100644 index 0000000..b491033 Binary files /dev/null and b/static/other/baker_poster_NER2021.pdf differ diff --git a/static/other/brockmeier_poster_distshift2021.pdf b/static/other/brockmeier_poster_distshift2021.pdf new file mode 100644 index 0000000..01b5c0e Binary files /dev/null and b/static/other/brockmeier_poster_distshift2021.pdf differ diff --git a/static/other/carlos_poster_NER2021.pdf b/static/other/carlos_poster_NER2021.pdf new file mode 100644 index 0000000..6074b73 Binary files /dev/null and b/static/other/carlos_poster_NER2021.pdf differ diff --git a/static/other/curtin_poster_summer2021.pdf b/static/other/curtin_poster_summer2021.pdf new file mode 100644 index 0000000..9b19df0 Binary files /dev/null and b/static/other/curtin_poster_summer2021.pdf differ diff --git a/static/other/graduatePdf2018.pdf b/static/other/graduatePdf2018.pdf old mode 100644 new mode 100755 diff --git a/static/other/karahan_poster_distshift2021.pdf b/static/other/karahan_poster_distshift2021.pdf new file mode 100644 index 0000000..3f1dd4b Binary files /dev/null and b/static/other/karahan_poster_distshift2021.pdf differ diff --git a/static/other/labombard_poster_summer2021.pdf b/static/other/labombard_poster_summer2021.pdf new file mode 100644 index 0000000..f1ca51a Binary files /dev/null and b/static/other/labombard_poster_summer2021.pdf differ diff --git a/static/papers/brockmeier2014_kernel_metric_learning.pdf b/static/papers/brockmeier2014_kernel_metric_learning.pdf new file mode 100644 index 0000000..c606167 Binary files /dev/null and b/static/papers/brockmeier2014_kernel_metric_learning.pdf differ diff --git a/static/papers/brockmeier2016_learning_recurrent_waveforms.pdf b/static/papers/brockmeier2016_learning_recurrent_waveforms.pdf new file mode 100644 index 0000000..0ee7695 Binary files /dev/null and b/static/papers/brockmeier2016_learning_recurrent_waveforms.pdf differ diff --git a/static/papers/brockmeier2019_improving_reference_prioritisation.pdf b/static/papers/brockmeier2019_improving_reference_prioritisation.pdf new file mode 100644 index 0000000..8db87ec Binary files /dev/null and b/static/papers/brockmeier2019_improving_reference_prioritisation.pdf differ diff --git a/static/papers/brockmeier2021_ML_for_documents.pdf b/static/papers/brockmeier2021_ML_for_documents.pdf new file mode 100644 index 0000000..d3856b6 Binary files /dev/null and b/static/papers/brockmeier2021_ML_for_documents.pdf differ