Output details
11 - Computer Science and Informatics
University of Plymouth
Top-Down Feedback in an HMAX-Like Cortical Model of Object Perception Based on Hierarchical Bayesian Networks and Belief Propagation
<24>This paper simultaneously extends the leading computational hierarchical model (HMAX) into two crucial directions:
top-down feedback and Bayesian interpretation. It also provides approximations of belief propagation that for the first time allow to simulate 30.000 nodes in a Bayesian model. Since November 2012 the paper has already been picked up by two reviews in top-end review journals (Trends Cognitive Science; Current Opinion Neurobiology), featuring very prominently in one of them. The algorithmic efficiency allows for large Bayesian real-world applications. The first author collaborates with the Applied Physics Lab (Johns Hopkins University) to use the proposed methods in sound recognition.