Implementation of Stochastic Adaptive Search with Hit-and-Run as a Generator
The Hit-and-Run algorithm iteratively generates a sequence of points in a set by taking steps of random length in randomly generated directions. Hit-and-Run is used as a sampling technique within a global optimization algorithm, as a generator in the context of simulated annealing. The ultimate goal is to approximate the desirable properties of stochastic adaptive search, in particular, the linear complexity of pure adaptive search. Such an algorithm is the Hit-and-Run version of simulated annealing with temperature equal to zero, called Improving Hit-and-Run because only points improving the objective function are accepted. The Hit-and-Run generator coupled with a Metropolis acceptance criterion is called Hide-and-Seek, and is motivated by theoretical results from adaptive search. Other variations of Hit-and-Run are also described here, as applied to global optimization.
KeywordsSimulated Annealing Global Optimization Feasible Region Coordinate Direction Global Optimization Problem
Unable to display preview. Download preview PDF.
- Boneh, A. and Golan, A. (1979). Constraints’ redundancy and feasible region boundedness by random feasible point generator. EURO I II: Third European Congress on Operations Research. Amsterdam, The Netherlands.Google Scholar
- Kiatsupaibul, S., Smith, R.L., and Zabinsky, Z.B. (2001). A discrete hit-and-run algorithm for generating samples from general multivariate distributions. Technical report, Department of Industrial and Opera- tions Engineering, The University of Michigan, Ann Arbor, Michigan.Google Scholar
- Kristinsdottir, B.P. (1997). Complexity Analysis of Random Search Al-gorithms. PhD thesis, University of Washington, Seattle, Washington.Google Scholar
- Murty, K.G. (1983). Linear Programming. John Wiley and Sons, New York, New York.Google Scholar
- Neogi, S. (1997). Design of Large Composite Structures using Global Optimization and Finite Element Analysis. PhD thesis, University of Washington, Seattle, Washington.Google Scholar
- Neogi, S., Zabinsky, Z.B., and Tuttle, M.E. (1994). Optimal design of composites using mixed discrete and continuous variables. In Proceedings of the ASME Winter Annual Meeting volume 52; Symposium on Processing, Design and Performance of Composite Materials, pages 91–107.Google Scholar
- Smith, R.L. (1984). Efficient Monte Carlo procedures for generating points uniformly distributed over bounded regions. Operations Research, 32 (1296–1308).Google Scholar
- Zabinsky, Z.B., Graesser, D.L., Tuttle, M.E., and Kim, G.I. (1992). Global optimization of composite laminate using Improving Hit-and-Run. In Floudas, C.A. and Pardalos, P.M., editors, Recent Advances in Glo-bal Optimization, pages 343–365. Princeton University Press, Princeton, New Jersey.Google Scholar
- Zabinsky, Z.B., Wood, G.R., Steel, M.A., and Baritompa, W.P. (1995). Pure adaptive search for finite global optimization. Mathematical Programming, 69 (443–448).Google Scholar