Abstract
The parallelization techniques used in the conventional computer technology need the overhead with respect to the synchronization cost. Moreover it is difficult to achieve a high speed- up ratio in general. On the other hand, turning to our brains, although they achieve high-speed information processing using only low-speed neurons, it is hard to imagine such parallelization takes place in our brains. Are there any simple and efficient parallelization techniques? In this paper we discuss about such parallelization technique for state pace search problem. We also prove a basic theorem about this parallelization, and show a simple guideline of using this technique. Although the neural network is inherently a massively parallel algorithm, we can realize the parallelization of neural networks, namely the parallelization of massively parallel, by using the proposed technique.
