ccorcos/deep-learning

Some deep learning models and experiments built with Theano.

Some Deep Learning models built with Theano

To Do

• save and reload model.

LSTM

• try on dice model

RNN

• tensor.alloc in RNN instead of repeat
• modify rnn to handle varying length sequences with a mask
• dropout?

MRNN

• character prediction
• hf optimization

MUSE

• USE with multiplicative units

Theano-users

• ask about rnn dropout
• ask about ubuntu install script

CNN

• conv net on images or maybe even imagenet

DA

• denoising autoencoder on mnist

Writing

• write about deep learning, https://imgur.com/a/Hqolp

TSNE

• low dimensional visualization: http://lvdmaaten.github.io/tsne/

Getting Started

Some examples use sparkprob to visualize probablity distributions at the commandline so you may need to install it

pip install sparkprob

All of the examples use the DL package. To use it:

cd DL
python setup.py develop

To unlink this package when you are done:

cd DL
python setup.py develop --uninstall

To load the datasets

cd datasets
curl -O http://www.iro.umontreal.ca/~lisa/deep/data/mnist/mnist.pkl.gz
curl -O http://www.iro.umontreal.ca/~lisa/deep/data/imdb.pkl