A general purpose neurocomputer

Abstract

Presents a neural network, composed of linear units with threshold, as the CPU of a stored program MIMD architecture. The Caianiello formalism, is introduced as an aid to implement the arithmetic and control algorithms, needed for the smooth running of this general-purpose system. That is, in the neural net both the arithmetic and logic algorithms and the operating system have been implemented. The latter is diffuse as it has been co-implemented with the single arithmetic operations. It controls each operation I/O, the input, output and intermediate data buffers, the clerical work associated to the beginning and the end of a task execution, etc. The neural net control is data-driven, i.e. the incoming data are the very signals telling the net to execute its task. As the net is data-driven, the system supports an efficient run time resource allocation algorithm. That is, at run time the incoming instructions chase the available resources and the waiting time, spent by the data in presence of idle resources, is minimized. At the same time, the system pipelines, automatically, nested loops, of arbitrary depth, and accepts unlimited recursive calls of routines.<>