DNN-Decide, A Novel Deep Neural Community (DNN) Based mostly Black-Field Optimization Framework For Analog Sizing
[ad_1]
This technical paper titled “DNN-Decide: An RL Impressed Optimization for Analog Circuit Sizing utilizing Deep Neural Networks” is co-authored from researchers at The College of Texas at Austin, Intel, College of Glasgow. The paper was a finest paper candidate at DAC 2021.
“On this paper, we current DNN-Decide, a novel Deep Neural Community (DNN) primarily based black-box optimization framework for analog sizing. Our methodology outperforms different black-box optimization strategies on small constructing blocks and enormous industrial circuits with considerably fewer simulations and higher efficiency. This paper’s key contributions are a novel sample-efficient two-stage deep studying optimization framework impressed by the actor-critic algorithms developed within the Reinforcement Studying (RL) group and its extension for industrial-scale circuits. That is the primary software of DNN primarily based circuit sizing on industrial scale circuits to one of the best of our information,” in keeping with the DAC presentation.
Discover the technical paper right here.
Authors: Ahmet F. Budak, Prateek Bhansali, Bo Liu, Nan Solar, David Z. Pan, Chandramouli V. Kashyap.
arXiv:2110.00211v1
Go to Semiconductor Engineering’s Technical Paper library right here and uncover many extra chip trade tutorial papers.
[ad_2]
Source_link
