Trust-Region Method with Deep Reinforcement Learning in Analog Design Space Exploration
Abstract
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.
Cite
@article{arxiv.2009.13772,
title = {Trust-Region Method with Deep Reinforcement Learning in Analog Design Space Exploration},
author = {Kai-En Yang and Chia-Yu Tsai and Hung-Hao Shen and Chen-Feng Chiang and Feng-Ming Tsai and Chung-An Wang and Yiju Ting and Chia-Shun Yeh and Chin-Tang Lai},
journal= {arXiv preprint arXiv:2009.13772},
year = {2021}
}
Comments
6 pages, 3 figures, 5 tables