Trust Region based MPS Method for Global Optimization of High Dimensional Design Problems

摘要

Mode Pursing Sampling (MPS) was developed as a global optimization algorithm for design problems involving expensive black-box functions. MPS has been found to be effective and efficient for problems of low dimensionality, i.e., the number of variables is less than 10. This work integrates the concept of the trust region into the MPS framework so that MPS can be applied to solve high dimensional optimization problems. Two trust regions are defined and their sizes are dynamically adjusted. Within each trust region, the search follows the traditional MPS process. Preliminary testing shows encouraging results.