Normalization and Similarity Measures
This appendix contains some additional information about normalization and similarity measures. As shown in Fig.4.1, the cognition module takes its input from the perception module. In practice, some aspects of normalization should be taken into consideration. Normalization depends on the properties of the data, as well as on the (di)similarity measures in the cognition module. In general, when the patterns are normalized beforehand, they should not be normalized during the matching process in SOM (6.2). For computational efficiency, it is advisable to adopt this strategy.
Unable to display preview. Download preview PDF.