(Professor of Computaional Neuroscience)
Wednesday 21st September 2011
B10 Seminar Room,
Alexandra House, 17 Queen Square, London, WC1N 3AR
How effective are neural model representations in capturing the higher-order correlations of natural images?
The physical structure of objects manifests itself in the higher-order correlations of natural images. The content of an image is still recognized easily if one removes all structure responsible for second-order correlation but otherwise leaves the higher-order correlations untouched. Since modeling higher-order correlations is a challenging task, it can be highly informative to examine to what extent neural model representation have the capacity to account for them. In this talk I will give an overview on the research done in my lab over the last 4-5 years in order to quantify the potential of different mechanisms such as orientation selectivity, contrast gain control, complex cell pooling or hierarchical architectures to capture these higher-order correlations. I will also present a new model---a Mixture of Conditional Gaussian Scale Mixtures---which achieves the best likelihood on natural images that has been reported to date. It also performs remarkably well as a texture synthesis algorithms demonstrating the perceptual relevance of the modeled correlations.