Abstract
Differential evolution (DE), one of evolutionary algorithms, was introduced by Stone and Price in 1995 as a population-based stochastic search technique for solving optimization problems in a continuous space. Individuals in DE are encoded as real valued vectors. Offspring are generated based on the difference vector between parents. In recent years, many extended versions of DE have been proposed to enhance the performance or robustness. In our research we aim to apply DE to optimization problems over discrete-valued domains. This paper propose an improved DE algorithm to solving Solving Quadratic Assignment Problems (QAP). A solution to a QAP can be represented by a permutation. Therefore, in the proposed method we use discrete valued coding. Individuals in DE are represented as discrete valued vectors. Also we propose the differential operator which adequately inherit characteristics of population to the offspring in discrete-valued domains. The performance of proposed DE is evaluated through experiments with benchmark problems of QAPLIB.
