Uncertainty aware Search Framework for Multi-Objective Bayesian Optimization with Constraints
This addresses the challenge of efficiently optimizing expensive-to-evaluate systems with multiple objectives and constraints, such as in circuit design, though it appears incremental as it builds on existing Bayesian optimization methods.
The paper tackles the problem of constrained multi-objective blackbox optimization with expensive evaluations by proposing the USeMOC framework, which reduces the number of simulations needed by over 90% to uncover optimized circuits in a voltage regulator design.
We consider the problem of constrained multi-objective (MO) blackbox optimization using expensive function evaluations, where the goal is to approximate the true Pareto set of solutions satisfying a set of constraints while minimizing the number of function evaluations. We propose a novel framework named Uncertainty-aware Search framework for Multi-Objective Optimization with Constraints (USeMOC) to efficiently select the sequence of inputs for evaluation to solve this problem. The selection method of USeMOC consists of solving a cheap constrained MO optimization problem via surrogate models of the true functions to identify the most promising candidates and picking the best candidate based on a measure of uncertainty. We applied this framework to optimize the design of a multi-output switched-capacitor voltage regulator via expensive simulations. Our experimental results show that USeMOC is able to achieve more than 90 % reduction in the number of simulations needed to uncover optimized circuits.