-
An Autoregressive Model is a model which performs regression on the previous values of a particular input.
More specifically, given a random variable which represents sequential data of the form
To predict
, we use as inputs in the regression model. -
A Latent Autoregressive Model is an autoregressive model where we maintain a summary
of the past observations for given sequential data . At the same time, we update
along with our prediction . The end result is that we can estimate
with and update of the form
-
One technique we can use is Windowing where, rather than using all of
as inputs, we simply use where is the size of the window. We do this so that our features have the same size.