There is a hideen layer L2_out = self.KG_l2(GRU_out_reshape) after GRU output in Architecture-1 in code, but which is not found in the paper.

[dys863]

As shown in the following lines, which is in the code KalmanNet_nn.py at lines 201-211, we can see there is a hidden layer L2_out = self.KG_l2(GRU_out_reshape), but this hidden layer is not found in the Fig.3 of the paper.

`
###########
### GRU ###
###########
GRU_in = torch.empty(self.seq_len_input, self.batch_size, self.input_dim).to(self.device,non_blocking = True)
GRU_in[0, 0, :] = La1_out
GRU_out, self.hn = self.rnn_GRU(GRU_in, self.hn)
GRU_out_reshape = torch.reshape(GRU_out, (1, self.hidden_dim))

    ####################
    ### Hidden Layer ###
    ####################
    L2_out = self.KG_l2(GRU_out_reshape)
    La2_out = self.KG_relu2(L2_out)

    ####################
    ### Output Layer ###
    ####################
    L3_out = self.KG_l3(La2_out)

`