Figure 5.

Total squared error between a typical data set and the exemplar matrix found by the SOM algorithm at each iteration, for different sets of parameters (number of exemplar nodes, learning rate [LR], or neighborhood contraction rate [NCR]). The default parameters (marked by an asterisk), which were implemented on the data, were 100 exemplar nodes, a learning rate of 0.1, and a neighborhood contraction rate of 0.05.