A Review of Field Computation
Bruce J. MacLennan
This report reviews the basic principles of field computation, a model of massively parallel analog computation, and discusses its applications in natural and artificial intelligence. We begin with the mathematical foundations and notation for field computation; Hilbert spaces provide the basic mathematical framework. Next we discuss examples of field computation in the brain, especially in its computational maps. Fields appear in a number of contexts, including activity at the axon hillocks, in patterns of axonal connection between areas, and in patterns of synaptic connection to dendrites. The following section presents examples of field computation in the brain and in other natural and artificial systems, including fields for sensorimotor processing, excitable media, and diffusion processes. Next we consider special topics in field computation in cognition, including the separation of information (semantics) from pragmatics, and the analysis of discrete symbols as field excitations. We also consider the relevance of universal mutivariate approximation theorems to general-purpose field computation. Then we discuss hardware specifically oriented toward field computation, including electronic, optical, and chemical technologies. Finally, we consider future directions for field computation research.
Published 2007-10-31 04:00:00 as ut-cs-07-606 (ID:128)