IPLAM: A High-Dimensional Expensive Simulation Optimization Method, With Application to Design Space Exploration in Processor

Design Space Exploration (DSE) in processors is an expensive discrete simulation optimization problem. The data requirements of the regular data-driven methods are so large that it is challenging to converge to a satisfactory solution within a limited simulation budget. Based on binary integer linear programming (BILP), an iteratively piecewise linear approximate method (IPLAM) is proposed for this kind of problems to reduce the dependence of simulation data. IPLAM starts from an initial reference point. Each iteration generates a set of trial points based on the most promising reference point by piecewise shift method. After evaluating the trial points, a local surrogate model is constructed for the unit neighborhood of the reference point. The surrogate model is then used to guide the exploration of the next reference point. In theory, IPLAM can converge to the global optimal point under mild assumptions, which is further verified by the numerical experiments. Meanwhile, the numerical experiments demonstrate that IPLAM outperforms the advanced Bayesian optimization and differential evolution methods on high-dimensional discrete closed-box test functions. Besides, the practical effectiveness of IPLAM is validated by an industrial case for processor DSE. Note to Practitioners—This paper proposes an efficient optimization method, IPLAM, for high-dimensional expensive simulation problems. To cope with the extremely limited simulation data, we construct the surrogate objective for the local space around the reference point by evaluating a set of trail points. The surrogate objective is then based on to locate the most promising point as the next reference point. We also provide theoretical guarantees of convergence for the proposed method. According to the numerical experiments, the proposed method demonstrates a much superior exploration efficiency and a lower simulation data requirement than competing methods. Besides, compared with an advanced Bayesian optimization method (HEBO) in an industrial case of processor DSE, the proposed method achieves a 47% improvement in convergence speed and a 15% reduction in simulation overhead while the difference of simulation score is marginal.