Content-Addressable Memory Storage by Neural Networks: A General Model and Global Liapunov Method

摘要

Many neural network models capable of content-addressable memory are shown to be special cases of the general model and global Liapunov function. These include examples of the additive, brain-state-in-a-box, McCulloch-Pitts, Boltzmann machine, shunting, masking field, bidirectional associative memory, Volterra-Lotka, Gilpin-Ayala, and Eigen-Schuster models. The Cohen-Grossberg model thus defines a general principle for the design of content addressable memory, that is shared by all model exemplars of such a general design constitutes a computational invariant. Such a general model and analytic method defines a computational framework within which specialized model exemplars may be compared to discover which models are best able to explain particular parametric data about brain and behavior, or to solve particular technological problems.