On Applications of Extreme Value Theory in Optimization
We present a statistical study of the distribution of the objective value of solutions (outcomes) obtained by stochastic optimizers, applied for continuous objective functions. We discuss the application of extreme value theory for the optimization procedures. A short review of the extreme value theory is presented to understand the investigations. In this chapter three optimization procedures are compared in this context: the random search and two evolution strategies. The outcomes of these optimizers applied to three objective functions are discussed in the context of extreme value theory and the performances of the procedures investigated, analytically and by simulations. In particular, we find that the estimated extreme value distributions and the fit to the outcomes characterize the performance of the optimizer in one single instance.
KeywordsWeibull Distribution Random Search Generalize Extreme Value Generalize Pareto Distribution Lower Tail
Unable to display preview. Download preview PDF.
- Bäck T, Hoffmeister F, Schwefel HP (1991) A survey of evolution strategies. In: Belew RK, Booker LB (eds) Genetic Algorithms: Proceedings of the Fourth International Conference at San Diego, Morgan Kaufmann, San Mateo, pp 2–9Google Scholar
- Fisher R, Tippett L (1928) Limiting forms of the frequency distribution in the largest particle size and smallest number of a sample. In: Proceedings of the Cambridge Philosophical Society, pp 180–190Google Scholar
- Hüsler J, Li D (2007) Testing extreme value conditions with applications. In: Reiss RD, Thomas M (eds) Statistical analysis of extreme values, 3rd edn, Birkhäuser, Basel, pp 144–151Google Scholar
- Reiss RD, Thomas M (1997) Statistical Analysis of Extreme Values with Application to Insurance, Finance, Hydrology and Other Fields. Birkhäuser, BaselGoogle Scholar