Comparative Study of Inhomogeneous Simulated Annealing Algorithms for Quadratic Assignment Problem

Abstract

Simulated annealing (SA)algorithm is a popular metaheuristic which has been widely used to solve many continuous and combinatorial optimization problems. Temperature cooling scheme is one of the key factors for its promising performance. Several studies have shown that reannealing can improve SA's performance on quadratic assignment problem (QAP). Using the benchmark instances from QAPLIB, this paper carried systematic experiments to compare the effect of rean-nealing on performance of inhomogeneous SA (ISA). Simulation results show that: (1)although reannealing can improve ISA's performance in general, it may has no effect on some type of QAP instances; (2)parameters for cooling schedule are not only sensitive to the type of QAP instance, but sensitive to the specific instance also; (3)the timing to activate the reannealing is sensitive to the type of QAP instances also. The source code is made publicly available as a benchmark for future comparisons.