SCS Undergraduate Thesis Topics

Ben Poole Tai Sing Lee Encoding Natural Priors in Neural Populations

Bayesian theories of the brain have provided insights into perception, but the underlying neural mechanisms which could implement these computations remains unknown. To perform Bayesian inference, sensory information must be combined with prior information about the natural world. We investigated how these natural priors could be learned and encoded in populations of neurons in primary visual cortex. We found that the distribution of neuronal tuning properties for depth-tuned neurons was very similar to the distribution of depths occurring in natural scenes. This finding is consistent with the hypothesis that neurons are performing optimal sampling of the natural environment based on the information maximization principle. By using the priors encoded in the tuning properties of neuronal populations, we were able to develop a framework for performing Bayesian inference in the brain.

Close this window