A Survey of Some Model-Based Methods for Global Optimization
Jiaqiao Hu, Yongqiang Wang, Enlu Zhou, Michael C. Fu, Steven I. Marcus · Systems & control · 2012
We review some recent developments of a class of random search methods: model-based methods for global optimization problems. Probability models are used to guide the construction of candidate solutions in model-based methods, which makes them easy to implement and applicable to problems with little structure. We have developed various frameworks for model-based algorithms to guide the updating of probabilistic models and to facilitate convergence proofs. Specific methods covered in this survey include model reference adaptive search, a particle-filtering approach, an evolutionary games approach, and a stochastic approximation-based gradient approach. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.