Probability Collectives using Response Surface estimation

This paper motivates and describes the use of Response Surface (RS) with Probability Collectives (PC) to handle large-scale optimization problems. The main characteristic of PC is that it optimizes the probability distribution of the variables rather than their values, thus different types of variables may be integrated into optimization procedure. The RS is used to approximate the utility evaluation of candidate solutions in PC. To improve the approximation accuracy, the Trust Region (TR) method is introduced to iteratively update the RS during optimization. Extensive simulations are conducted to demonstrate the effectiveness of the proposed algorithm.