struct GRUCell {
&FloatTensor W_ir;
&FloatTensor W_hr;
&FloatTensor b_ir;
&FloatTensor b_hr;
&FloatTensor W_iz;
&FloatTensor W_hz;
&FloatTensor b_iz;
&FloatTensor b_hz;
&FloatTensor W_in;
&FloatTensor W_hn;
&FloatTensor b_in;
&FloatTensor b_hn;
}r_t = sigmoid(x_t@W_ir + h@W_hr + b_ir + b_hr) reset gate z_t = sigmoid(x_t@W_iz + h@W_hz + b_iz + b_hz) update gate n_t = tanh(x_t@W_in + b_in + r_t * (h@W_hn + b_hn)) candidate h_t = (1 - z_t) * n_t + z_t * h_{t-1}
Fields
Methods
Parameterized (module.h): every weight tensor.
= drop()