arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2009.13772cs.LGcs.AI
收藏

模拟设计空间探索中基于深度强化学习的信赖域方法

Trust-Region Method with Deep Reinforcement Learning in Analog Design Space Exploration

National Tsing Hua University(台湾清华大学);MediaTek Inc.(联发科)

Kai-En Yang, Chia-Yu Tsai, Hung-Hao Shen, Chen-Feng Chiang, Feng-Ming Tsai, Chung-An Wang, Yiju Ting, Chia-Shun Yeh, Chin-Tang Lai

摘要:

本文介绍了模拟设计空间搜索的新观点。为了最小化上市时间,这种努力比现有技术中定义的全局优化更好地被视为约束满足问题。我们将基于模型的代理,对比无模型学习,实现信任区域的战略。因此,简单的前馈网络可以用监督学习来训练,其中收敛相对微不足道。实验结果表明,数量级的改善搜索迭代。此外,适应PVT条件的前所未有的考虑。在台积电5/6 nm工艺的电路上,我们的方法实现了超越人类设计师的性能。此外,该框架已在工业环境中生产。

英文摘要:

This paper introduces new perspectives on analog design space search. To minimize the time-to-market, this endeavor better cast as constraint satisfaction problem than global optimization defined in prior arts. We incorporate model-based agents, contrasted with model-free learning, to implement a trust-region strategy. As such, simple feed-forward networks can be trained with supervised learning, where the convergence is relatively trivial. Experiment results demonstrate orders of magnitude improvement on search iterations. Additionally, the unprecedented consideration of PVT conditions are accommodated. On circuits with TSMC 5/6nm process, our method achieve performance surpassing human designers. Furthermore, this framework is in production in industrial settings.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑