junjiehu/deepy

A highly extensible deep learning framework

deepy: A highly extensible deep learning framework based on Theano

deepy is a deep learning framework for designing models with complex architectures.

Many important components such as LSTM and Batch Normalization are implemented inside.

Although highly flexible, deepy maintains a clean high-level interface.

From deepy 0.2.0, you can easily design very complex computational graphs such as Neural Turing Machines.

Example codes will be added shortly.

Dependencies

• Python 2.7 (Better on Linux)
• numpy
• theano
• scipy for L-BFGS and CG optimization

Tutorials (Work in progress)

http://deepy.readthedocs.org/en/latest/

Clean interface

# A multi-layer model with dropout for MNIST task.
from deepy import *

model = NeuralClassifier(input_dim=28*28)
model.stack(Dense(256, 'relu'),
            Dropout(0.2),
            Dense(256, 'relu'),
            Dropout(0.2),
            Dense(10, 'linear'),
            Softmax())

trainer = MomentumTrainer(model)

annealer = LearningRateAnnealer(trainer)

mnist = MiniBatches(MnistDataset(), batch_size=20)

trainer.run(mnist, controllers=[annealer])

Examples

Enviroment setting

• CPU

source bin/cpu_env.sh

• GPU

source bin/gpu_env.sh

MNIST Handwriting task

• Simple MLP

python experiments/mnist/mlp.py

• MLP with dropout

python experiments/mnist/mlp_dropout.py

• MLP with PReLU and dropout

python experiments/mnist/mlp_prelu_dropout.py

• Maxout network

python experiments/mnist/mlp_maxout.py

• Deep convolution

python experiments/mnist/deep_convolution.py

• Elastic distortion

python experiments/mnist/mlp_elastic_distortion.py

• Recurrent visual attention model
  • Result visualization

python experiments/attention_models/baseline.py

Variational auto-encoders

• Train a model

python experiments/variational_autoencoder/train_vae.py

• Visualization the output when varying the 2-dimension latent variable

python experiments/variational_autoencoder/visualize_vae.py

• Result of visualization

Language model

Penn Treebank benchmark

• Baseline RNNLM (Full-output layer)

python experiments/lm/baseline_rnnlm.py

• Class-based RNNLM

python experiments/lm/class_based_rnnlm.py

• LSTM based LM (Full-output layer)

python experiments/lm/lstm_rnnlm.py

Char-based language models

• Char-based LM with LSTM

python experiments/lm/char_lstm.py

• Char-based LM with Deep RNN

python experiments/lm/char_rnn.py

Deep Q learning

• Start server

pip install Flask-SocketIO
python experiments/deep_qlearning/server.py

• Open this address in browser

http://localhost:5003

Auto encoders

• Recurrent NN based auto-encoder

python experiments/auto_encoders/rnn_auto_encoder.py

• Recursive auto-encoder

python experiments/auto_encoders/recursive_auto_encoder.py

Train with CG and L-BFGS

• CG

python experiments/scipy_training/mnist_cg.py

• L-BFGS

python experiments/scipy_training/mnist_lbfgs.py

Other experiments

DRAW

See https://github.com/uaca/deepy-draw

# Train the model
python mnist_training.py
# Create animation
python animation.py experiments/draw/mnist1.gz

Highway networks

• http://arxiv.org/abs/1505.00387

python experiments/highway_networks/mnist_baseline.py
python experiments/highway_networks/mnist_highway.py

Effect of different initialization schemes

python experiments/initialization_schemes/gaussian.py
python experiments/initialization_schemes/uniform.py
python experiments/initialization_schemes/xavier_glorot.py
python experiments/initialization_schemes/kaiming_he.py

---

Sorry for that deepy is not well documented currently, but the framework is designed in the spirit of simplicity and readability. This will be improved if someone requires.

Raphael Shu, 2016