Design Space Exploration

Design space exploration (DSE) aims to efficiently identify optimal configurations within a vast set of possible designs, addressing challenges in various fields from hardware synthesis to AI model deployment. Current research heavily utilizes machine learning, particularly deep neural networks, Bayesian optimization, reinforcement learning, and generative adversarial networks, to create surrogate models for faster and more accurate exploration of complex design spaces, often incorporating graph neural networks for representing and analyzing design structures. This accelerates the design process, improves the quality of resulting designs, and enables more informed decision-making across diverse applications, ranging from optimizing hardware architectures for specific tasks to enhancing the efficiency and fairness of AI systems.