Memory Models for Improving Probabilistic Tabu Search for Continuous Real Variables

This paper proposes that current memory models in us for tabu search algorithms are at best evolving, as opposed to adaptive, and that significant improvement can be made by considering the structure of human memory. By introducing a new memory structure, intended to reduce the number of evaluations required at each step, it is possible to allow the search method to learn about the solution space in which it is operating. The memory model is based on the transfer of events from episodic memory into generalised rules stored in semantic memory. By adopting such a model, the search algorithm can intelligently explore the solution space in response to what has been learned to date and update the stored knowledge as the search progresses. This implements a truly adaptive memory for the method.