I would like to train multivariate time series models with time varying weight information (time-varying relationship between data and labels). My understanding is that for example, autoregressive models and LSTM models have only one type of weight information learned from training data which is not what I am intended.
What I am looking for is that I would like to understand potential multiple types of relationships between time series data and time series labels (for example, from time TA to TB, we found y=f(x), and from time TB to TC, we found y=g(x), and overall, we found f(x), g(x), j(x), k(x), and when these functions are switching.
Could you tell me with which models, we can accomplish this?
