codekansas / theano-rnn

:mag: Some RNN implementations in Theano

theano-rnn

Language modeling stuff has been moved to a new repo (it didn't really match with this one anymore, and I wanted to get my commit rate up).

Some RNN implementations in Theano, along with various optimizers. In particular:

• vanilla.py: Basic RNN implementation
• gru.py: Gated recurrent unit
• lstm.py: Long short term memory
• optimizers.py: rmsprop, sgd, and momentum (including Nesterov momentum)

Each file has a framework for generating random data and fitting the RNN model to that data, so you can play around with it.