Multinomial bayesian learning for modeling classical and nonclassical receptive field properties

Authors:
Haruo Hosoya
Affiliations:
-
Venue:
Neural Computation
Year:
2012

Citing 8
Cited 1

Learning invariance from transformation sequences

Neural Computation
Connectionist learning of belief networks

Artificial Intelligence
Simple-cell-like receptive fields maximize temporal coherence in natural video

Neural Computation
Perceptual grouping and the interactions between visual cortical areas

Neural Networks - 2004 Special issue Vision and brain
Pattern Recognition and Machine Learning (Information Science and Statistics)

Pattern Recognition and Machine Learning (Information Science and Statistics)
Computational theory and applications of a filling-in process at the blind spot

Neural Networks
Bayesian interpretation of border-ownership signals in early visual cortex

ICONIP'10 Proceedings of the 17th international conference on Neural information processing: theory and algorithms - Volume Part I
Unsupervised learning of generative and discriminative weights encoding elementary image components in a predictive coding model of cortical function

Neural Computation

A computational model of motor areas based on bayesian networks and most probable explanations

ICANN'12 Proceedings of the 22nd international conference on Artificial Neural Networks and Machine Learning - Volume Part I

Quantified Score

Hi-index	0.00

Visualization

Abstract

We study the interplay of Bayesian inference and natural image learning in a hierarchical vision system, in relation to the response properties of early visual cortex. We particularly focus on a Bayesian network with multinomial variables that can represent discrete feature spaces similar to hypercolumns combining minicolumns, enforce sparsity of activation to learn efficient representations, and explain divisive normalization. We demonstrate that maximal-likelihood learning using sampling-based Bayesian inference gives rise to classical receptive field properties similar to V1 simple cells and V2 cells, while inference performed on the trained network yields nonclassical context-dependent response properties such as cross-orientation suppression and filling in. Comparison with known physiological properties reveals some qualitative and quantitative similarities.