GTSOM: Game Theoretic Self-Organizing Maps

Self-Organizing Maps (SOM) is a powerful tool for clustering and discovering patterns in data. Input vectors are compared to neuron weight vectors to form the SOM structure. An update of a neuron only benefits part of the feature map, which can be thought of as a local optimization problem. A global optimization model could improve representation to data by a SOM. Game Theory is adopted to analyze multiple criteria instead of a single criteria distance measurement. A new training model GTSOM is introduced to take into account cluster quality measurements and dynamically modified learning rates to ensure improved quality.