Abstract
The simulated annealing is an optimization algorithm, by means of which near-optimal solutions can be searched in a solution space. The solution space is defined by a set of neighbor relations between solutions and a schedule of the temperature. Thus how to construct and determine them is a key to the success of application of the algorithm. However, it has not been studied in detail how to effectively analyze and eval-uate them. In this paper, we propose a new method to visualize a solution space explored in the algorithm. First, a transition probability between two solutions, which cannot be expressed in the existing methods, is considered. Its visual expression is also considered. Next, based on transition probabilities, a “space plot” can be obtained. It can reflect the behavior of the algorithm at a given temperature, and indicate the quality of a given set of neighbor relations. Several examples of space plots are also shown to demonstrate the effectiveness of the proposed method.
