hsg@romeo.cs.duke.edu (Henry Greenside) (09/21/89)
Does anyone know of any references comparing the performance of a Hopfield network for associative memory with nearest-neighbor look up? What I am after is whether typical basins of attractions for fixed points defined by the Hopfield model are well approximated by the regions of space closest to the fixed points (Voronoy polyhedra). If this is the case, no dynamics is necessary to do associative memory: as soon as one knows which basin of attraction you are in, you know the answer. Henry Greenside