Presentation
Graph Machine Learning assisted Analog Circuit Designing
DescriptionThis paper presents a graph-based semi-supervised learning framework to accelerate the design cycle of analog circuits. It develops a new way of graph construction by leveraging the homophily within circuit and its instances. An integration of traditional deep-learning methods with cutting-edge graph neural networks (GNNs) helps to achieve better results with fewer labelled examples. The proposed method called efficient analog sizing through constrained optimization (EASCO), summarizes the prediction task over unlabelled instances and incorporates constrained optimization to achieve an optimal design. The simulated performance of different circuits using derived parameters in 180nm CMOS process confirms the efficacy of proposed algorithms.
Event Type
Work-in-Progress Poster
TimeTuesday, July 11th6:00pm - 7:00pm PDT
LocationLevel 2 Lobby
AI
Autonomous Systems
Cloud
Design
EDA
Embedded Systems
RISC-V
Security
