An improved Harmony Search Algorithm embedded with a novel piecewise opposition based learning algorithm

Abstract

Harmony Search Algorithm (HS) is one of the most popular, music-inspired meta-heuristic algorithm. Since its conception HS has been used to solve many complex problems. However, this population based algorithm and its variants suffer from slow convergence speed to the globally optimal solutions. Hence they are computationally expensive. Opposition Based Learning Theory, a machine learning algorithm addresses this issue by considering both estimates and counter-estimates i.e. guess and counter-guess of a candidate solution, population and opposite population of a population based algorithm etc. simultaneously. Although this approach shows great promise, the problem of slow convergence rate in Harmony Search Algorithm is still not completely alleviated. We introduce some improvements over the Opposition based Learning Theory to accelerate the convergence rate of such algorithms. The proposed scheme employs a piecewise counter estimate updating technique while computing a candidate solution. In the present work, the proposed Opposition based Learning technique has been embedded in the framework of Harmony Search Algorithm. An exhaustive set of test functions is used in the experimental setup. The results obtained from the experiments are very promising.