Research Manuscript: From machine learning to graphs: a new wave towards analog design
Event TypeResearch Manuscript
Analog Design, Simulation, Verification and Test
TimeWednesday, July 13th1:30pm - 3:00pm PDT
Location3004, Level 3
DescriptionThis session presents the new EDA techniques used in analog and mixed-signal design. Graph-based sparsification is presented to speed up power grid transient simulations. Two-step reinforcement learning is introduced to accelerate DC circuit simulation. Circuit sizing is boosted by fine-tuning layout neural models. Finally, a Gaussian process classification and varying-sigma sampling approach is used for yield optimization of analog circuits.