Reinforcement Learning Approach in Electronic Design Automation Earns Top Honors at DAC 2023
Aug 31, 2023 —
Researchers from the Georgia Tech Computer-Aided Design (GTCAD) Laboratory in the School of Electrical and Computer Engineering have received this year’s Design Automation Conference (DAC) Best Paper Award For Research.
The award-winning paper was co-authored by Motorola Solutions Foundation Professor Sung Kyu Lim and Yi-Chen Lu (ECE Ph.D. ’23, currently at Apple), in collaboration with a team from Synopsys, Inc comprised of Wei-Ting Chan, Deyuan Guo, Vishal Khandelwal, and Sudipto Kundu.
The research, titled “RL-CCD: Concurrent Clock and Data Optimization using Attention-Based Self-Supervised Reinforcement Learning,” received Best Paper recognition out of 1,157 submissions. It presents a Reinforcement Learning (RL) agent in Concurrent Clock and Data (CCD) optimization — a technique used in modern computer design tools to improve the performance and reliability of digital circuits. The introduction of an RL agent enables systems to intelligently enhance their ability to correctly rank violating endpoints according to machine learning-based optimization strategies. This contributes to an optimization flow that maximizes the overall efficiency and effectiveness of the system's performance.
The team was presented the award at DAC 2023 — the flagship conference in electronic design automation (EDA) — in San Francisco this July.
Last year, Lim and his research team were presented with the Donald O. Pederson Best Paper Award for their research on Compact-2D physical design tools at DAC. The work was recognized as the best paper published in IEEE’s Transactions on Computer-Aided Design of Integrated Circuits and Systems (IEEE TCAD), the flagship journal of the IEEE Council on Electronic Design Automation (CEDA).
Top photo caption: The team receiving the Best Paper Award For Research this July at the Design Automation Conference. Left to right: Sung Kyu Lim, Wei-Ting Chan, Yi-Chen Lu, and Deyuan Guo (not pictured: Vishal Khandelwal, and Sudipto Kundu).