Toolbox Documentation

Yann toolbox is divided into different parts and written in many files. Below is an index of all files and the API documentations.

• network - The network module
• layers - Contains the definitions of all the types of layers.
  • abstract - abstraction class
  • input - input layer classes
  • fully_connected - fully connected layer classes
  • conv_pool - conv pool layer classes
  • merge - merge layer classes
  • flatten - flatten layer classes
  • output - output layer classes
  • random - random layer classes
  • transform - transform layers
    • LICENSE
  • batch_norm - Batch normalization layer classes
• modules - Modules that are external to the network which the network uses
  • optimizer - optimizer class
  • datastream - datastream class
  • visualizer - visualizer class
  • resultor - resultor class
• core - Core module contains the fundamentals operations of the toolbox
  • conv - Definitions for all convolution functions.
  • pool - Definitions for all Pooling functions.
  • activations - Definitions for all activations functions.
  • errors - Definitions for all point-wise error functions.
  • operators - Definitions for all basic operators written for yann .
• utils - utilities that can be used as enhancement for the toolbox’s functionality
  • dataset - provides a nice port to benchmark and matlab-based datasets
  • graph - provides a nice port to networkx methods related to Yann
  • pickle - provides a way to save the network’ parameters as a pickle file.
  • raster - provides a visualization for rasterizing images.
    • LICENSE
• special - contains tools for special types of networks
  • gan - provides a inherited network class for a gan network.
  • datasets - provides quick methods to produce common datasets.