CS205: Building with Artificial Intelligence

Hill climbing can be improved by randomly accepting a sub-optimal move under specific circumstances. What probability would you accept? The suboptimal move turns out to be complex and connected to many factors. These factors are balanced by the simulated annealing algorithm. Early in the search process, suboptimal moves are accepted with higher probability, and as the process winds down, suboptimal moves are not as readily accepted. In the long run, simulated annealing produces the best likelihood of finding global optima relative to more naive algorithms. Even so, finding the global optima in all cases is not guaranteed.