How generative models develop in predictive processing

Abstract

In the predictive processing theory generative models are used to make predictions about future inputs, only processing the unpredicted parts: the prediction error. In order to be able to make increasingly better predictions, predictions errors should be used to update and revise the generative models. How this model generation and revision is established is still unclear. This research focuses on comparing a basic- and a more specified generative model to find out if this is a possible way that the generative models develop. This research will be done using Lego Mindstorms EV3-robots.