抄録
A novel Markov process which is a normalized version of a Simulated Annealing(SA) process is considered. The process, which is called NSA process, converges to a limit distribution which has a pointed peek at an optimum solution of the energy function. The convergence rate is sufficiently large, even under such an extremely low "temperature"that NSA yeilds a probabilistically extended neighborhood search. A simulation algorithm of NSA is described, and some numerical experiments on comparison between SA and NSA are made. The results of the experiments show that NSA takes effect in global optimization.
