Optimization of Simulation via Quasi-Newton Methods

This paper discusses the application of quasi-Newton methods to optimization of simulation. Specifically, it describes a general methodology that combines response surface methodology and other optimization techniques with quasi-Newton methods. Using quasi-Newton methods in the vicinity of the optimum speeds up the convergence rate of response surface methodology, the gradient, and search methods. From the class of quasi-Newton methods, we recommend BFGS and Davidon's methods. These methods are effective without exact line searches. Two examples demonstrate the effectiveness of the proposed methodology. INFORMS Journal on Computing, ISSN 1091-9856, was published as ORSA Journal on Computing from 1989 to 1995 under ISSN 0899-1499.