Home - dmlc/cxxnet GitHub Wiki

This is the documentation for cxxnet, a short and fast implementation for neural network in CPU/GPU

Getting started

• Read Tutorial to get started, play cxxnet on MNIST, run a feed forward neural network and convolution neural network in 3 minutes!

=

Parameter Setting

To understand more about cxxnet setting, this section will introduce all parameters of cxxnet and the effect of parameters. In general, cxxnet configuration file contains 4 kinds of configurations in a single file:

• Global Setting

• Set global parameters for neural network, related to device selection, running control.

• Data Input Setting
  • Set input data configurations.
• Updater Setting

• Set parameters(learning rate, momentum) for learning procedure

• Layer Setting
  • Configure network, and setup each layers.

=

Guide to train ImageNet

• Read the Guide to start train your own ImageNet Model.
• We have released pretrianed model on ImageNet 2010 Dataset

=

Code Layout

• core: core implementation of layers and updaters of neural net
• nnet: implementation of network that connects layers and updaters together
• io: handles data input and preprecessing
• utils: some utils to help simple tasks