Toolbox include different types of neural nets:
- MLP (Multilayer Perceptron which can be trained by the backpropagation algorithm)
- Generative RBM (Restricted Boltzmann Machine used for unsupervised learning)
- Classification RBM (Restricted Boltzmann Machine used for the classification task)
All training methods use next improvements:
- Nesterov's gradient
- local learning rates
- global learning rate strategy