Skip to content

Latest commit

 

History

History
20 lines (15 loc) · 695 Bytes

README.md

File metadata and controls

20 lines (15 loc) · 695 Bytes

sybilsystem

Construction on indefinite hiatus

A MATLAB library for prototyping deep neural networks (that can be trained using backpropagation), named for the eponymous "supercomputer". Developed for clear code, not speed. To get started run the initialize script.

This work has been based on the Unsupervised Feature Learning and Deep Learning Tutorial and utilises some of the helper functions provided.

Wishlist

  • Classification results
  • Batch accuracy
  • Convolution and pooling layers (only FC now)
  • Exporting networks and layers
  • Automatic saving during training
  • Visualisation of learnt features