PC-Opt：基于分区和征服的多代理复杂模拟电路优化器.pdf

1、Pohang University of Science and TechnologyDepartment of Electrical EngineeringCAD and SoC Design Lab.Tel.+82-54-279-2883Web.http:/csdl.postech.ac.krCONTACTPC-Opt:Partition and Conquest-based Optimizer using Multi-Agents for Complex Analog CircuitsYoungchang Choi,Sejin Park,Ho-Jin Lee,Kyongsu Lee,Ja

2、e-Yoon Sim,and Seokhyong Kang2025.01.22ASIA SOUTH PACIFIC DESIGN AUTOMATED CONFERENCE 2025OutlineI.IntroductionII.Proposed MethodIII.Experimental Setup and ResultsIV.Conclusion*Appendix,References.I.IntroductionChallenges for Analog Circuit Design Traditionally relies on human expertise3ExpertOh my

3、god!As technology of transistors becomes more advanced,sizing circuits becomes more difficult.I.IntroductionApplication of Artificial Intelligence(AI)Application of AI to sizing AMS circuits Bayesian Optimization(BO)1 Reinforcement Learning(RL)2,3,4 Previous research typically overlooks the optimiza

4、tion of complex analog circuits.4Bayesian OptimizationReinforcement Learning WEIBO,TCAS-I,2018 1 AutoCkt,DATE,2020 2 GCN-RL,DAC,2020 3 DNN-Opt,DAC,2021 41 W.Lyu et al.,“An efficient bayesian optimization approach for automated optimization of analog circuits”,TCAS-I,2017.2 K.Settaluri et al.,“Autock

5、t:Deep reinforcement learning of analog circuit designs”,DATE,2020.3 H.Wang et al.,“Gcn-rl circuit designer:Transferable transistor sizing with graph neural networks and reinforcement learning”,DAC,2020.4 C.Ding et al.,“Dnn-opt:An RL inspired optimization for analog circuit sizing using deep neural

6、networks”,DAC,2021.I.IntroductionOptimizing complex analog circuits Applying MADDPG framework 5 handle this issue.States,actions,and rewards are shared.DDPG requires extensive simulations,increasing overall optimization time.55 Zhang et al.,“Automated design of complex analog circuits with multiagen