Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Electrical and Computer Engineering

First Advisor

Subhash C. Kak


This dissertation puts forth an original theory of cortical neural processing that is unique in its view of the interplay of chaotic and stable oscillatory neurodynamics and is meant to stimulate new ideas in artificial neural network modeling. Our theory is the first to suggest two new purposes for chaotic neurodynamics: (i) as a natural means of representing the uncertainty in the outcome of performed tasks, such as memory retrieval or classification, and (ii) as an automatic way of producing an economic representation of distributed information. We developed new models, to better understand how the cerebral cortex processes information, which led to our theory. Common to these models is a neuron interaction function that alternates between excitatory and inhibitory neighborhoods. Our theory allows characteristics of the input environment to influence the structural development of the cortex. We view low intensity chaotic activity as the a priori uncertain base condition of the cortex, resulting from the interaction of a multitude of stronger potential responses. Data, distinguishing one response from many others, drives bifurcations back toward the direction of less complex (stable) behavior. Stability appears as temporary bubble-like clusters within the boundaries of cortical columns and begins to propagate through frequency sensitive and non-specific neurons. But this is limited by destabilizing long-path connections. An original model of the post-natal development of ocular dominance columns in the striate cortex is presented and compared to autoradiographic images from the literature with good matching results. Finally, experiments are shown to favor computed update order over traditional approaches for better performance of the pattern completion process.