My webpage will also contain the work I’m doing to investigate the generalization ability of neural networks, specifically for the multi-modal case.

Please see the corresponding git repo

Currently, this repo contains the work I’ve done to train Large Margin Deep Networks. Large Margins are important because of robustness to adversarial perturbations, robustness to noisy training data, and increased generalizability. The following are papers surrounding this subject:

  1. Large Margin Deep Networks for Classification
  2. On the depth of deep neural networks, a theoretical view
  3. Robust Large Margin Deep Neural Networks