CELLULAR NEURAL NETWORKS AS A GENERAL MASSIVELY-PARALLEL COMPUTATIONAL PARADIGM

In this paper is presented the use of the discrete-time cellular neura l network (DTCNN) paradigm to develop algorithms devised for general-p urpose massively parallel processing (MPP) systems. This paradigm is d efined in discrete N-dimensional spaces (lattices) and is characterize d by the locality of the direct information transmission between the s pace points (cells) and by continuous values of data and parameters; t he DTCNN paradigm is thus able to express most of the;typical MPP appl ications. A general version of a DTCNN has been implemented and optimi zed for three MPP architectures, namely the Connection Machines CM-2 a nd CM-5 and the Gray T3D. The comparison between the three machine per formances with those achieved by a standard SPARC-20 workstation shows that, particularly with large lattices, the speed-up allowed in the c omputational times is significant and the range of solvable problem si zes is widely extended.