Multi-core CPU implementation of deep learning for 2D and 3D convolutional networks (ConvNets).

Currently we only support linux environments.

make
make clean

If compile is successful, an executalbe named znn will be generated under the directory bin.

An executable will be generated here.

Matlab functions for preparing training data and analyzing training results.

C++ source code.

• core -- core classes for constructing ConvNets and performing multi-core parallelized computations.
• cost_fn -- cost (objective) functions for training ConvNets.
• error_fn -- linear and/or non-linear activation functions for neurons.
• front_end -- an interface for specifying ConvNet architecure, training data, and training options.
• initializer -- random initializers for weights of ConvNets.

General purpose C++ library, written and maintained by Aleksander Zlateski.

• Recursive Training of 2D-3D Convolutional Networks for Neuronal Boundary Detection

• Core Architecture
• How to Tutorial

• Aleksander Zlateski <zlateski@mit.edu>
• Kisuk Lee <kisuklee@mit.edu>