Stochastic co-derivative of harmony search algorithm

Abstract

A novel stochastic co-derivative was developed based on the density information of multiple vectors in the harmony search (HS) algorithm. While the existing stochastic partial derivative represents the probability with which certain candidate value is selected when searching for a new vector, this co-derivative represents how much one variable in a vector is correlated with other variables. The proposed stochastic co-derivative harmony search (CDHS) algorithm was applied to the design of hydraulic structure, and found better results than the original harmony search algorithm in terms of the number of reaching the global optimum and the number of function evaluations. The algorithm was also tested with a bigger hydraulic problem, finding better solutions in terms of least and average costs when compared with other phenomenon-mimicking algorithms such as genetic algorithm (GA), simulated annealing (SA), and tabu search (TS).