As we discussed in the previous section, there are multiple ways of architecting a network to generate sequences of outputs. In this section, we will learn about the encoder decoder way of generating outputs, and also about the one-to-one mapping of inputs to outputs network on a toy dataset so that we have a strong understanding of how this works.
Let's define a sequence of inputs and a corresponding sequence of outputs, as follows (the code file is available as Return_state_and_sequences_working_details.ipynb in GitHub):
input_data = np.array([[1,2],[3,4]])output_data = np.array([[3,4],[5,6]])
We can see that there are two time steps in an input and that there is a corresponding output to the ...