- Support for embeddings_layer
- Support for group_layer
- Support for merge_layer
- Support for rnn_layer (Recurrent Neural Network (RNN))
- Support for lstm_layer (Long Short Term Memory (LSTM))
- Support for pretty printing and live printing
- Faster Batch Normalization
- GPU Support for dropout
- GPU Support for shuffle
Initial version (was rolling released before) with the following main features:
- RBM and CRBM support
- Deep Belief Network
- Contrastive Divergence
- Stochastic Gradient Descent
- Artificial Neural Networks
- Convolutional Neural Networks
- Auto-Encoders
- Pooling
- Dropout
- Batch Normalization
- Adaptive Learning Rates
- Basic data augmentation
- Datasets support: MNIST, CIFAR-10, ImageNet
- High CPU and GPU performance