A CONCURRENT ADAPTIVE CONJUGATE-GRADIENT LEARNING ALGORITHM ON MIMD SHARED-MEMORY MACHINES

A concurrent adaptive conjugate gradient learning algorithm has been d eveloped for training of multilayer feed-forward neural networks and i mplemented in C on a MIMD shared-memory machine (CRAY Y-MP/8-864 super computer). The learning algorithm has been applied to the domain of im age recognition. The performance of the algorithm has been evaluated b y applying it to two large-scale training examples with 2,304 training instances. The concurrent adaptive neural networks algorithm has supe rior convergence property compared with the concurrent momentum back-p ropagation algorithm. A maximum speedup of about 7.9 is achieved using eight processors for a large network with 4,160 links as a result of microtasking only. When vectorization is combined with microtasking, a maximum speedup of about 44 is realized using eight processors.