Fig. 3
From: Biomedical semantic indexing by deep neural network with multi-task learning
The unit of GRU. \( {z}_t^j \)is an update gate, where the superscriptjindicates the update gate is the jthelement in the update gate vectorzt, the subscript tindicates the time step t. \( {r}_t^j \)denotes a reset gate. \( {h}_{t-1}^j \) represents a hidden state of the time step t ‐ 1, \( {h}_t^j \) represents a hidden state of the time step t and \( {\overset{\sim }{h}}_t^j \)is a candidate activation