Predictive processing: testing a definition of stabilizing average prediction error
Keywords
No Thumbnail Available
Authors
Issue Date
2019-06-19
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
The predictive processing theory postulates that the brain functions as a hierarchical
prediction machine with the goal of minimizing future prediction error. Kwisthout et al.
(2017) have formalized the predictive processing theory in terms of categorical probability
distributions. In this thesis I propose that stabilizing average prediction error could be used as
a measure to indicate whether the probabilities in a generative model defined in terms of the
formalization of Kwisthout et al. could indicate real-world probabilities. I tested a by me
introduced definition of stabilizing average prediction error in a robot experiment. The
definition was not successful. The probabilities in the generative model were stabilized
according to the definition before they actually were. The scope of this thesis was too small to
make informed choices about the details of the definition, and the fluctuation of mean
prediction error was not take into account in the definition An alternative definition is
proposed in which the latter is resolved. Future work could be to let the variables in the
definition be dependent on factors in the generative model rather than the variables having set
values.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen