Page 255 - Classification Parameter Estimation & State Estimation An Engg Approach Using MATLAB

244 UNSUPERVISED LEARNING

(a) (b)
No of iterations = 0 No of iterations = 10

(c) (d)
No of iterations = 25 No of iterations = 100

Figure 7.10 The development of a one-dimensional self-organizing map, trained on
a two-dimensional uniform distribution: (a) initialization; (b)–(d) after 10, 25 and
100 iterations, respectively

This introduces two additional scale parameters which have to be set by
the user.
Although the SOM offers a very flexible and powerful tool for map-
ping a data set to one or two dimensions, the user is required to make
many important parameter choices: the dimension of the grid, the num-
ber of neurons in the grid, the shape and initial width of the neighbour-
hood function, the initial learning rate and the iteration dependencies of
the neighbourhood function and learning rate. In many cases a two-
dimensional grid is chosen for visualization purposes, but it might not fit