抄録
In Cellular Neural Network (CNN), a design of neighborhood or output function has a great influence on the efficiency of self-recall. In this research, we focused on improving the efficiency of CNN by designing the neighborhood. In CNN which has a local connectivity, it is an important index in improving the efficiency how we design the neighborhood of each cell. In this paper, we proposed a new method of designing the neighborhood, which reduces the recall time while maintaining the classification capability. Next, simulations using model patterns were demonstrated, and it was shown that the proposed method was more effective than the conventional method. Furthermore, we applied the CNN designed by the proposed method to the abnormal sounds diagnosing, and it was shown that the proposed method was also effective in actual application.
