Abstract. Experiments of neural network training procedure parallelization are conducted. Several styles of parallelization are described and compared, estimations of neural network size and training set size that allow speedup on two-processors SMP machine are obtained.
Victor G. Tsaregorodtsev