An artificial neural network accelerator using general purpose 24 bit floating point digital signal processors

Abstract

An artificial neural network (ANN) accelerator named Neuro Turbo was implemented using four recently developed general-purpose 24-b floating-point digital signal processors (DSP) MB86220. The Neuro Turbo is a MIMD (multiple-instruction, multiple-data) parallel processor having four ring-coupled DSPs and four dual-port memories (DPM). It is designed compactly to plug into the extender slots of the NEC personal computer PC98 series. The performance was evaluated by constructing a neural network to recognize the 26 type fonts of the alphabet set. Processing speeds of 2 MCPS (million connections per second) for the learning procedure and 11 MCPS for the forward pass were achieved.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>