A New Parallel Search Algorithm for Non-Linear Function Optimization

Abstrac/-This paper proposes a new parallel search algorithm using evolutionary programming and quasi-simplex technique (EPQS). EPQS produces the offspring from three ways in parallel: 1) Using the Gaussian mutation, 2) Using the Cauchy mutation, and 3) Using the quasi-simplex techniques. The quasi-simplex technique uses the ideal of classical simplex technique and produces four prospective individuals by using the reflection, expansion and compression operations. EPQS selects the parents for the next generation from all the parents and offspring. EPQS takes the diversity of offerings into consideration by generating the offspring from as many as possible ways while it maintains a substantial convergence rate. Experimental studies on six typical benchmark functions have shown that the proposed algorithm is more effective than the competing algorithms.