Speaker: Mr Jonathan Hunt, Computational Neuroscience Laboratory, Queensland Brain Institute
Title: Patterns, preferences and the primary visual cortex
The world in which we live is filled with a fascinating mixture of structured patterns and apparent randomness. By taking advantage of patterns, biological systems can reduce the computational load of decoding the state of the world. We have attempted to examine the statistical structure of the visual world and the ways in which this structure influences the development of the primary visual cortex and is utilized in visual processing.
Edges are an important component of natural scenes. Edge arrangement in natural scenes contains statistical patterns such as co-circularity. Using binocular rivalry, we have found that the visual system has a "preference" for certain edge arrangements. Surprisingly, we have found that primary visual cortex maps and lateral connectivity do not appear to be strongly influenced by these statistics. However, primary visual cortex receptive fields are determined almost entirely by natural scene statistics. We have demonstrated this by showing that a simple unsupervised learning model postdicts primary visual cortex receptive field development in both normal and modified rearing conditions. Overall, we have found the primary visual cortex is deeply influenced by the statistics of visual input, but not always in the ways one might first expect.
Seminars will be followed by a sandwich lunch for all attendees.
Your attendance at these seminars is both welcomed and encouraged.
For a list of upcoming seminars at QBI, go to: