PyTorch Highway Networks

Highway networks implemented in PyTorch.

Highway Equation

Just the MNIST example from PyTorch hacked to work with Highway layers.

Todo

  • Make the Highway nn.Module reuseable and configurable.
  • Why does softmax work better than sigmoid? This shouldn’t be the case…
  • Make training graphs on the MNIST dataset.
  • Add convolutional highway networks.
  • Add recurrent highway networks.
  • Experiment with convolutional highway networks for character embeddings.

Notes

  • ELU doesn’t work better than RELU for the layer activation.
  • Softmax seems to work better than sigmoid for the gate function?!

GitHub

https://github.com/c0nn3r/pytorch_highway_networks