 |
Center for Nonlinear Studies |
Abstract and unifying theoretical models based on conservation principles, such as energy and mass, have contributed immensely to the understanding of diverse phenomena in Physics. Sensory processing regions of the brain (e.g. thalamus, olfactory bulb, sensory cortex) have a massive amount of re-entrant top-down feedback pathways which regulate bottom-up processing. Such close regulation is suggestive of conservation principles. However, pathways with large numbers of feedback connections are difficult to analyze because the mathematics becomes highly nonlinear. Thus a unified theory is necessary. The computational basis of conservation of information is that an input should not pass more information than is justified to the next layer. Thus inputs are regulated by the outputs they activate. Subsequently, each input’s contribution (i.e. salience) is adjusted through feedback regulation by its associated outputs. This model displays unparalleled performance given simultaneous patterns. It also inherently performs and displays: (1) Cognitive phenomena, (2) Pattern grouping, (3) Decisions given ambiguous or impoverished patterns, and (4) Plasticity within sensory processing. Through the conservation of information principle diverse phenomena can be captured by a unified model. With a better appreciation of the link between feedback processes and conservation properties, it is possible attain more robust and unified understanding of sensory recognition processing.
Host: Luis Bettencourt, T-5