Example $4\mathrm{.}6$ A SOM to cluster letters from different fonts: linear structure A linear structure $($ with R $=1)$ gives a better distribution of the patterns onto the available cluster units. The winning node J and its topological neighbors $($ J$+1$ and J $1)$ are allowed to learn on each iteration. Note that in general, the neighboring nodes that learn do not initially have weight vectors that are particularly close to the input pattern.

UNIT PATTERNS UNIT PATTERNS

$6$ k$220$ c$1,$c$2,$c$3$

$10$ j$1,$j$2,$j$322$ d$2$

$14$ e$1,$e$223$ b$2,$e$3$

$16$ k$1,$k$325$ a$1,$a$2,$a$3$

$18$ b$1,$b$3,$d$1,$d$3$