Package ai.djl.nn.recurrent

Class GRU

java.lang.Object
ai.djl.nn.AbstractBaseBlock
ai.djl.nn.AbstractBlock
ai.djl.nn.recurrent.RecurrentBlock
ai.djl.nn.recurrent.GRU
All Implemented Interfaces:
Block
public class GRU extends RecurrentBlock
GRU is an abstract implementation of recurrent neural networks which applies GRU (Gated Recurrent Unit) recurrent layer to input.

Current implementation refers the [paper](http://arxiv.org/abs/1406.1078) - Gated Recurrent Unit. The definition of GRU here is slightly different from the paper but compatible with CUDNN.

The GRU operator is formulated as below:

$$ \begin{split}\begin{array}{ll} r_t = \mathrm{sigmoid}(W_{ir} x_t + b_{ir} + W_{hr} h_{(t-1)} + b_{hr}) \\ z_t = \mathrm{sigmoid}(W_{iz} x_t + b_{iz} + W_{hz} h_{(t-1)} + b_{hz}) \\ n_t = \tanh(W_{in} x_t + b_{in} + r_t * (W_{hn} h_{(t-1)}+ b_{hn})) \\ h_t = (1 - z_t) * n_t + z_t * h_{(t-1)} \\ \end{array}\end{split} $$