The investigations described in this application are designed to link established results from visual physiology and psychophysics with well- known principles of information processing, and, in the process establish the extent that such principles underlie the processing of sensory inputs. This approach employs analytic and computational techniques whose predictions are tested by established physiologic data. Portions of the visual cortex are among the most well-studied regions of the brain, and such studies are complemented by extensive bodies of psychophysical experiments. Consequently, the properties of individual neurons and visual channels are well-established, and the characteristic arrangement of these neurons over the visual cortex is known with increasing refinement. There is growing focus on the connections between these different neurons, and the interactions mediated by these connections. However, the question whose answer lags behind all of this progress is, why? It is intuitively appealing to think that the properties of the individual neurons, their spatial arrangement over the cortex, and their interactions are all geared towards representing visual stimuli in a manner that maximizes the use of available neural resources. Such resources would include the number of neurons, the size of their synapses, and the length of the connections between them. This application continues work that quantitatively develops and tests these concepts. Thus far, work of this type has revealed how many of the cardinal properties of visual neurons and their arrangement over the cortex emerges when the synapse is the fundamental unit of neural information processing. The current application continues its pursuit of these concepts by seeking an explanation for the connections between the many types of visual neurons and the interactions ascribed to these connections. It is hypothesized that such connections permit the visual system to continue to optimize itself in a changing visual environment, and that many complex phenomena produced by these interactions are the consequence of the optimization. The results from these studies will be used to guide the development of more complex models of neural information processing, and may be applicable to other regions of the central nervous system. In addition, this material provides an important quantitative link between biologic and synthetic methods of image representation.
| Gottschalk, Allan; Sexton, Matthew G; Roschke, Guilherme (2004) Multiplicative neural noise can favor an independent components representation of sensory input. Network 15:291-311 |
| Gottschalk, Allan (2002) Derivation of the visual contrast response function by maximizing information rate. Neural Comput 14:527-42 |
