Bayesian Columnar Networks for Grounded Cognitive Systems

There are two elementary building blocks whose intimate relationship has been overlooked so far: The cortical column as ubiquitous local processor underlying all cognitive operations and the Bayesian framework with the generic operation of calculating posterior probabilities given likelihoods and prior probabilities. In this contribution we try to bridge this gap by providing a detailed mapping between parts of the cortical column and processing steps involved in Bayesian reasoning. The resulting columnar network is implemented and fully integrated within an intentional vision system, which is also shortly characterized.