How to use neural networks to take a scalar as an input and gives a time series as an output

Hello there everyone

I have this mind-bugging question about neural networks. we normally use equisampled neural networks (for instance: we give 5 features each of which has 100 samples and we get a feature as an output with 100 sample). How about we train a neural network that, for example, takes 2 features each of which has 5 samples and gives out a time series.

Suppose that we have a differential equation as follows:

dy/dt = at/b

where a and b are constants. we solve this equation five times for different values of a and b and get a time series each time (so we have 5 time series). how can I train a neural network in matlab that by entering an arbitrary a and b, get a time series? Thanks in advance for your time devoted to this question.