Learning and Categorization in Modular Neural Networks

This text introduces a new neural network model, called CALM, for categorization and learning in neural networks. The internal structure of a CALM module is derived from the neocortical minicolumn, and a pivotal psychological concept in the CALM algorithm is self-induced arousal. The author demonstrates how this model can learn the world superiority effect for letter recognition, and he discusses a series of studies which simulates experiments in implicit and explicit memory, involving amnesic experiments. Pathological, but psychologically accurate behaviour is produced by "lesioning" the arousal system of these models. The author also introduces as an illustrative practical application a small model that learns to recognize handwritten digits. The book contains an introduction to generic algorithms, a computing method based on the biological metaphor of evolution, and it is demonstrated how these generic algorithms can be used to design network architectures with superior performance. The role of modularity in parallel hardware and software implementations is discussed in some depth.
Several types of implementation are considered, including transputer networks and a dedicated 400-processor neurocomputer built by the developers of CALM. The book ends with an evaluation of the psychological and biological plausibility of CALM models and a general discussion of generalization, interference and representational capacity of modular neural networks. This book should be of interest to researchers in psychology, cognitive science, neuroscience, computer simulation sciences, parallel computer architectures, pattern recognition, and to anyone interested in neural networks, neurocomputers and neurosimulators.