Includes:

• Restricted Boltzmann Machine (RBM)
  • Gaussian-Bernoulli unit
  • NRLU hidden unit
  • Softmax hidden unit
  • Persistent Contrastive divergence
  • Dropout for hidden unit
• Deep Belief Network (DBN)
• RNN-RBM
  • independent RBM pre-training
• Variational LSTM auto-encoder
• Quasi Reccurent NN

These extensions were validated in Keras 1.2.0 and python 3.5 (Anaconda). Tensorflow and Theano backend also work fine in Keras. Please refer to some examples to use libraries.

Reference:

• RBM, DBN : https://github.com/wuaalb/keras_extensions

• RNNRBM : http://deeplearning.net/tutorial/rnnrbm.html

• Variational LSTM Auto-Encoder : https://github.com/jayhack/LSTMVRAE

• Quasi Reccurent NN https://arxiv.org/pdf/1611.01576v1.pdf