cxxnet

CXXNET (spelled as: C plus plus net) is a neural network toolkit build on mshadow(https://github.com/tqchen/mshadow). It is yet another implementation of (convolutional) neural network. It is easily configured via config file, and can get the state of art performance.

People: Tianqi Chen, Naiyan Wang, Mu Li, Bing Xu

Documentation and Tutorial

Features

• Small but sharp knife: the core part of the implementation is less than 2000 lines
• Based on parameter-server, cxxnet supports multi-GPU training and distributed training with elegant speed.
• Build with mshadow, a tensor template library for unified CPU/GPU computation. All the functions are only implemented once, as a result. cxxnet is easy to be extended by writing tensor expressions.
• Python/Matlab interface for training and prediction.

Build Guide

cxxnet is designed to require less third party library. The minimal requirement is MKL/CBLAS/OpenBLAS and MShadow(which can be downloaded automatically). Other dependence can be set by editing make/config.mk before make.

• For users who want train neural network in less time, we suggest you buy a NVIDIA cuda-enabled video card and install CUDA in your system, then set USE_CUDA = 1 in make/config.mk to enable GPU training.
• For users who want to better speed up on convolution neural network, we suggest you install CuDNN R2 and set USE_CUDNN=1 in make/config.mk.
• For users who want to train on images, libjpeg or libjpeg-turbo is required for decoding images. We suggest you install OpenCV and set USE_OPENCV=1 to enable augmentation iterator.
• For MKL users who want to use Python interface, we suggest you change MShadow make config file to link to MKL in static way.