Adaptive annealing for chaotic optimization

The chaotic simulated annealing algorithm for combinatorial optimization problems is examined in the light of the global bifurcation structure of the chaotic neural networks. We show that the result of the chaotic simulated annealing algorithm is primarily dependent upon the global bifurcation structure of the chaotic neural networks and unlike the stochastic simulated annealing infinitely slow chaotic annealing does not necessarily provide an optimum result. As an improved algorithm, the adaptive chaotic simulated annealing algorithm is introduced. Using several instances of 20- and 40-city traveling salesman problems, efficiency of the adaptive algorithm is demonstrated.