Deep Markov Model (DMM) for time series prediction

Hi everyone,

I want to use DMM for time series prediction. My training data is x[1:T], and I want to predict x[T:T+t_forecast]. However, it seems that the DMM guide model requires observed data x[1:T+t_forecast] to make inference for the variational distribution q(z[1:T+t_forecast] | x[1:T+t_forecast]). This suggests that the current example in DMM might not be applicable for prediction problems, or perhaps I am misunderstanding something. Could someone help clarify this or suggest a way to handle this for prediction applications?

Thank you!

to do prediction forward in time you would use the guide to sample latents z up to when you have data (e.g. time T) and then use the model to go forward in time using e.g. GatedTransition and Emitter

1 Like

Dear Martin,

Super! Thank you for your clarification.