Some applications of simulated annealing to pattern recognition
Lei Xu · 2003
The simulated annealing technique for solving combinatorial problems is applied to: cluster analysis, isomorphisms of attributed relational graphs, piecewise curve fitting, and feature selection. A novel class of clustering algorithms based on simulated annealing are presented. One such algorithm, ALKMEANS, is proposed as a contrast to the commonly used heuristic clustering algorithm KMEANS; test results demonstrate that ALKMEANS is superior to KMEANS. A simulated annealing algorithm, ALISON, is presented for the problem of isomorphisms of relational graphs.>