Hybrid optimization-an experimental study

Abstract

This paper compares the performance of a hybrid optimization method to that of pure gradient based methods. Our hybrid optimization method comprises of an initial adaptive ordinal search phase followed by a gradient ascent (descent) phase. The adaptive ordinal search phase consists of fixing the size of the design population and ranking the members of the population using an estimated value of the performance. Members of the design population for the next stage are picked using the top designs of the previous population. This process is achieved via a variation on the standard genetic algorithm [l].Recently it was shown that ranks of populations are relatively insensitive to simulation noise[2], as the experimental data show, this fact is useful in using short simulation runs to improve the search efficiency before the onset of the final gradient ascent (descent) phase.

Stochastic Approximation with Simulated Annealing as an Approach to Global Discrete-Event Simulation Optimization

“…SA/SAN exhibits the first-order Markovian property, as the (i+1) st control is a function of the i th control alone, ignoring information collected previously in the search. Garai, Ho, and Sreenivas (1992) discuss this limitation. Another potential extension would be to augment SA/SAN with intelligent global search heuristics, such as a neural network overlay in and/or tabu search, in imitation of OptQuest.…”