# neural network stuff

not unlike [my dsp repo,](https://github.com/notwa/dsp)
this is a bunch of half-baked python code that's kinda handy.
i give no guarantee anything provided here is correct.

don't expect commits, docs, or comments to be any verbose.

### heads up

this was formerly a gist.
**i might rewrite the git history**
at some point to add meaningful commit messages.

## other stuff

if you're coming here from Google: sorry, keep searching.
i know Google sometimes likes to give random repositories a high search ranking.
maybe consider one of the following:

* [keras](https://github.com/fchollet/keras)
  for easy tensor-optimized networks.
  strong [tensorflow](http://tensorflow.org) integration as of version 2.0.
  also check out the
  [keras-contrib](https://github.com/farizrahman4u/keras-contrib)
  library for more components based on recent papers.
* [theano's source code](https://github.com/Theano/theano/blob/master/theano/tensor/nnet/nnet.py)
  contains pure numpy test methods to reference against.
* [minpy](https://github.com/dmlc/minpy)
  for tensor-powered numpy routines and automatic differentiation.
* [autograd](https://github.com/HIPS/autograd)
  for automatic differentiation without tensors.

## dependencies

python 3.5+

numpy scipy h5py sklearn dotmap

## contributing

i'm just throwing this code out there,
so i don't actually expect anyone to contribute,
*but* if you do find a blatant issue,
maybe [yell at me on twitter.](https://twitter.com/antiformant)