Programming a massively parallel, computation universal system: Static behavior

Publisher Website
Google Scholar
Abstract
Massively parallel systems are presently the focus of intense interest for a variety of reasons. A key problem is how to control, or ‘‘program’’ these systems. In previous work by the authors, the ‘‘optimum finding’’ properties of Hopfield neural nets were applied to the nets themselves to create a ‘‘neural compiler.’’ This was done in such a way that the problem of programming the attractors of one neural net (called the Slave net) was expressed as an optimization problem that was in turn solved by a second neural net (the Master net). The procedure is effective and efficient. In this series of papers we extend that approach to programming nets that contain interneurons (sometimes called ‘‘hidden neurons’’), and thus we deal with nets capable of universal computation. Our work is closely related to recent work of Rummelhart et al. (also Parker, and LeChun), which may be viewed as a special case of this formalism and therefore of ‘‘computing with attractors.’’ In later papers in this series, we present the theory for programming time dependent behavior, and consider practical implementations. One may expect numerous applications in view of the computation universality of these networks.

This publication has 0 references indexed in Scilit: