Recurrent Neural Networks

This module belongs to the category " Deep learning algorithms" .

Description

This module is used to initiate a Recurrent Neural Networks estimator based on SmartPredict library.

Parameters

The Recurrent Neural Networks module

Optimization parameters

The available optimizers for the RNN are:

• adam

• rmsprop

• adagrad

• adamax

• sgd

Other optimization parameters can also be set: Learning rate , Beta Gradient, Clipping by Norm or by Value, Epsilon Momentum, Rho.

Loss function

The loss function can be:

• binary_xentropy

• categorical_xentropy

• sparse_categorical _xentropy

• hinge

• mae

• cosine

The Recurrent Neural Network module contains a range of loss functions .

The module also contains a range of parameters such as : layer cells, timestamp, hidden recurrent units...