Visual Stability Improvement of SOM's Feature Map by Initial Value Assignment

MOMOI Shinji
MIYOSHI Tsutomu

Abstract -- In SOM learning, learning result depends on initial value of 
feature map and the location of the node or the distance between nodes on 
feature map is important factor to determine feature of individual data. For 
example, in data detection of hematopoietic tumors, given data is detected by 
location of the feature map. In this paper, we forcused on visual stability of 
SOM feature map, and we proposed new initialization method of SOM feature map. 
The purposes of proposed method are improvement of visual stability of SOM 
feature map, and utilization of generalization ability of SOM. By experiments, 
proposed method is visually stable than conventional method in the point of 
feature map location, and the computational complexity of proposed method is 
greatly reduced.

Keywords-component; self-organizing map; feature maps; visual stability;improvement method;