3D SOM Initialization Pattern Dictionary Algorithm Based on FCM Clustering
Ruihua Dong, Xueyan Zhang, Hongsong Li · Journal of Physics Conference Series · 2019
Abstract Aiming at the problem that the three-dimensional SOM reconstruction effect of the traditional pattern dictionary initialization algorithm is sensitive to the input order of the pattern, a three-dimensional SOM initialization mode dictionary algorithm based on FCM clustering is proposed. Calculate the mean square error of the training vector set and use the FCM algorithm to aggregate the resulting mean square differences into three categories. The average values are arranged in ascending order, and a certain pattern is extracted in the training vector at the same interval to form an initial pattern dictionary. The experimental results show that the 3D SOM initialization mode dictionary algorithm based on FCM clustering reduces the search time, increases the source matching degree, and improves the overall performance of the 3D SOM algorithm.