Advances in Multi-Objective Optimization and Circuit Placement

The field of multi-objective optimization is moving towards more efficient and effective methods for solving complex problems. Recent developments have focused on improving the diversity of Pareto optimal solutions and enhancing hypervolume performance. Additionally, there is a growing interest in applying black-box optimization techniques to chip placement tasks, with a focus on developing unified benchmarks and evaluating the effectiveness of different algorithms. Noteworthy papers include MOBO-OSD, which proposes a novel multi-objective Bayesian optimization algorithm, and BBOPlace-Bench, which introduces a benchmark for evaluating black-box optimization algorithms for chip placement tasks. Amortized Active Generation of Pareto Sets is also notable for its introduction of a new framework for online discrete black-box multi-objective optimization.

Sources

MOBO-OSD: Batch Multi-Objective Bayesian Optimization via Orthogonal Search Directions

Amortized Active Generation of Pareto Sets

Accelerating Electrostatics-based Global Placement with Enhanced FFT Computation

Tree-Cotree-Based IETI-DP for Eddy Current Problems in Time-Domain

BBOPlace-Bench: Benchmarking Black-Box Optimization for Chip Placement

Omnipresent Yet Overlooked: Heat Kernels in Combinatorial Bayesian Optimization

Built with on top of