Hello Thonn,
It seems you're looking to conduct single-input, multi-output sequence training. To accomplish this, you'll need to construct a custom training loop that incorporates automatic differentiation. You can design a neural network in such a way that its output at the current time step (k) serves as the input for the subsequent time step (k+1). This design allows a single input to generate a series of outputs over time. Sentence generation is a common application of this technique.
Unfortunately, MATLAB does not offer built-in functionality for this specific process, so you'll need to implement your own custom training loop. I recommend reviewing the provided documentation for further guidance on this topic,
To clarify, you'll need to structure your model loss so that the output from the initial forward pass is fed back as the input for the subsequent pass, continuing this process until the entire sequence is generated. The rest of the custom training process, including loss calculation and gradient updates based on the computed loss, remains consistent with standard procedures.
Hope this helps.
