NGAFID-SSL: Self-Supervised Learning with General Aviation Data

Link to paper.

Team

Name	Email
Aidan LaBella	aidan_labella@brown.edu
Aditya Iyer	aditya_iyer@brown.edu
Charlie Duong	charles_duong@brown.edu
Elise Carman	elise_carman@brown.edu
Nathan DePiero	nathan_depiero@brown.edu
Justin Long	pak_iong_long@brown.edu

Resources

Visuals

• Circuit Diagram of LOC-I Calculations

Videos

• Aidan's presentation to ACM SAC '22 on NGAFID data

Code

• NGAFID Codebase
• TST Codebase
• SimCLR Codebase

Papers

• Loss of Control/Stall Index
• Predictive Maintenance
• Phase of Flight ID with SOMs
• (Class Paper) Noisy Timeseries SSL
• Guillotine Regularization

Installation

$ conda env create --name ngafid-ssl --file ngafid_ssl_environment.yml
$ conda activate ngafid-ssl
$ python run.py

Config file

To change running configurations, pass keyword arguments to the run.py file.

$ python run.py -data ./datasets --dataset-name NGAFID --log-every-n-steps 100 --epochs 100 

To run on CPU use the --disable-cuda option. For 16-bit precision GPU training, use the --fp16_precision flag. This will use Pytorch built in AMP training.