neopentane / mnist-dnn

Deep neural network for handling MNIST image recognition

MNIST Deep Neural Network

A deep neural network for MNIST image recognition with the following key features:

• supports unlimited number of layers, nodes and weights (only restriction is memory)
• supports fully connected and convolutional layers
• supports following activation functions: SIGMOID, TANH, RELU
• light weight architecture with a very small memory footprint
• super fast! :-)

Compile and run source code

The repository comes with a pre-configured makefile. You can compile the source simply by typing

$ make

in the project directory. The binary will be created inside the /bin folder and can be executed via

$ ./bin/mnist-dnn

Code Review

If you're interested in how the code works take a look at my blog entry where I review the code for this deep neueral network in detail.

• Code Review for Deep Neural Network for MNIST Handwriting Recognition

Documentation

The /doc folder contains a doxygen configuration file. When you run it with doxygen it will create updated HTML documentation in the /doc/html folder.

• Doxygen Code Documentation

MNIST Database

The /data folder contains the original MNIST database files.

For more informaton on MNIST see Yann Lecun's THE MNIST DATABASE of handwritten digits

Version 1.0

Published: February 2016