基于Theano的可擴展深度學習框架deepy。

Dependencies

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

Clean interface

# MNIST Multi-layer model with dropout.
from deepy.dataset import MnistDataset, MiniBatches
from deepy.networks import NeuralClassifier
from deepy.layers import Dense, Softmax, Dropout
from deepy.trainers import MomentumTrainer, LearningRateAnnealer

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])</pre>

Examples 

Enviroment setting 


CPU 
</ul>

source bin/cpu_env.sh
 

GPU 
</ul>

source bin/gpu_env.sh
 
MNIST Handwriting task 


Simple MLP 
</ul>

python experiments/mnist/mlp.py
 

MLP with dropout 
</ul>

python experiments/mnist/mlp_dropout.py
 

MLP with PReLU and dropout 
</ul>

python experiments/mnist/mlp_prelu_dropout.py
 

Deep convolution 
</ul>

python experiments/mnist/deep_convolution.py
 

Elastic distortion 
</ul>

python experiments/mnist/mlp_elastic_distortion.py
 

Recurrent visual attention model 

Result visualization 
</ul>
</li>
</ul>

python experiments/attention_models/baseline.py
 
Language model 

Penn Treebank benchmark 


Baseline RNNLM (Full-output layer) 
</ul>

python experiments/lm/baseline_rnnlm.py
 

Class-based RNNLM 
</ul>

python experiments/lm/class_based_rnnlm.py
 

LSTM based LM (Full-output layer) 
</ul>

python experiments/lm/lstm_rnnlm.py
 
Char-based language models 


Char-based LM with LSTM 
</ul>

python experiments/lm/char_lstm.py
 

Char-based LM with Deep RNN 
</ul>

python experiments/lm/char_rnn.py
 
Deep Q learning 


Start server 
</ul>

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

Open this address in browser 
</ul>

http://localhost:5003
 
Auto encoders 


Recurrent NN based auto-encoder 
</ul>

python experiments/auto_encoders/rnn_auto_encoder.py
 

Recursive auto-encoder 
</ul>

python experiments/auto_encoders/recursive_auto_encoder.py
 
Train with CG and L-BFGS 


CG 
</ul>

python experiments/scipy_training/mnist_cg.py
 

L-BFGS 
</ul>

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
 
Other features 

Auto gradient correction 


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. 




項目主頁：http://www.baiduhome.net/lib/view/home/1434971044591