深度循环神经网络H t 1 = f 1 ( H t − 1 1 , X t ) ⋮ H t j = f j ( H t − 1 j , H t j − 1 ) ⋮ O t = g ( H t L ) \begin{aligned} \mathbf{H}_t^1 &= f_1(\mathbf{H}_{t-1}^1, \mathbf{X}_t) \\ &\vdots \\ \mathbf{H}_t^j &= f_j(\mathbf{H}_{t-1}^j, \mathbf{H}_t^{j-1}) \\ &\vdots \\ \