Deducing Principles in Natural Perception
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact David MacKay.
What are the underlying computational principles that biology uses to transform the raw sensory signal into a hierarchy of representations that subserve higher-level perceptual tasks? One hypothesis is that biological representations are optimal from the viewpoint of statistical information processing, and are adapted to the statistics of the natural sensory environment. In the initial stages of sensory coding, information has to be coded in a way that makes best use of the available resources. I will show how the optimal solution to coding natural images with a population of noisy neurons predicts many properties of retinal coding. I will also show that the same approach can be applied to natural sounds to explain many aspects of the auditory code at the cochlear nerve. Finally, I will present work that extends this general principle to higher levels of visual processing. I will show that efficient representations of the statistics of local image regions can form stable, invariant representations of edges, contours, and textures. These results also provide a novel functional explanation for non-linear effects of complex cells in the primary visual cortex.
This talk is part of the Inference Group series.
This talk is included in these lists:
Note that ex-directory lists are not shown.
|