Randomized Optimization
Dirk P. Kroese, Thomas Taimre, Zdravko I. Botev · Wiley series in probability and statistics · 2011
This chapter discusses optimization methods that have randomness as a core ingredient. Such randomized algorithms can be useful for solving optimization problems with many local optima and complicated constraints, possibly involving a mix of continuous and discrete variables. Randomized algorithms are also used to solve noisy optimization problems, in which the objective function is unknown and has to be obtained via Monte Carlo simulation. The chapter considers randomized optimization methods for both noisy and deterministic problems, including stochastic approximation, the stochastic counterpart method, simulated annealing, evolutionary algorithms, and the cross-entropy method. Controlled Vocabulary Terms cross-entropy method; Monte Carlo methods; Stochastic approximation; stochastic processes