Open vicgalle opened 5 years ago
Currently, we are only using the previous timestep for the input of the convolutional model. We should be able to stack a LSTM or GRU cell after the convolutional filters to learn long-term dependencies.
Currently, we are only using the previous timestep for the input of the convolutional model. We should be able to stack a LSTM or GRU cell after the convolutional filters to learn long-term dependencies.