2002 Volume 38 Issue 1 Pages 97-103
In this paper, for variable hierarchical structure learning automata with S-model stationary random environment at each level, a new definition of optimal path is proposed based on the arithmetic mean of average rewards, and an LR-I type learning algorithm is constructed. The learning propertiy of our algorithm is considered theoretically, and it is proved that the probability of finding the optimal path can approach 1 as mach as possible by using our algorithm. In numerical simulations, the usefulness of our algorithm is shown.