A visualization of walking around the latent space of an Autoencoder trained on MNIST.
Contributors:
- Milo Chang
- Jamie Tang
- Simon Socolow
For example, you can see what the model thinks is in between an 8 and a 9: 
python3 UI.py will start the demo, which loads the pre-trained model from newmodel.pt and ten digits from firstTenDigits.pt.
The code for the model can be found in model.py, and uses an architecture inspired by this notebook.
Notes:
- If you start at a 1 and go towards a 3, it won't look very good. But if you take a few steps towards a 4, then go to a 3, you can get a much better 3.
Check out our notebook to learn more about Autoencoders and a hands-on denoising application!
