/auto-grad

Toy Automatic Differentiation Library

Primary LanguagePythonMIT LicenseMIT

Auto-Grad

Toy Python implementation of automatic differentiation.

  • Includes both forward and reverse modes of AD.
  • Example of a 2 layer Neural Network in examples/simple_nn.py.
  • Test coverage of many methods.
  • SGD, SGD with Momentum, SGD with Nesterov Momentum avaliable.

Installation

Run: python setup.py install

Examples

See tests/ and examples/ for how to use.

Run $ python examples/simple_nn.py to train a 2 layer neural network on MNIST.