Optimization using statistical models

Project level: PhD, Masters, Honours

The field of meta-heuristic optimization offers a variety of powerful methods for solving difficult optimization problems, and requires no prior knowledge other than the cost function and any constraints. These methods include simulated annealing, genetic algorithms, estimation of distribution algorithms and the cross-entropy method. The latter two methods build statistical models to describe the lower-cost regions of the state space, and aim to move towards the global optimum as the algorithm iterates. Projects in this area will seek to find solutions to high-dimensional optimization problems. A challenge is to find to find highly effective statistical models for most optimization problems and understand the relationship between these models and other steps in the algorithm, such as selection and mutation. Such methods can be applied to a wide variety of real-world optimization problems. Reference:  Gallagher M. Wood I. Keith J. Sofronov G. (2007) Bayesian Inference in Estimation of Distribution Algorithms Proceedings of the IEEE Congress on Evolutionary Computation 127-133 Singapore IEEE Press.