Friday, Feb. 21, 2025
10:20 – 11:10 a.m. (CST)
ETB 1020
Yu Wang
PhD Candidate, Dept. of Computer Engineering
University of California, Santa Barbara
Title: “Enable Efficient Bayesian Optimization with Semi-Supervised Learning and In-Context Learning of Large Language Models”
Abstract
Bayesian Optimization (BO) is a powerful framework for optimizing expensive black-box functions, but its efficiency is often hindered by two key challenges: (1) difficulties in encoding domain knowledge and (2) data scarcity that limits surrogate model generalization. In this talk, I will present the recent efforts from our group to address these challenges.
First, we introduce ADO-LLM, which leverages the domain priors of Large Language Models (LLMs) to assist BO in tackling the challenging analog circuit sizing problem. By proposing feasible and high-potential design parameters, LLMs enable BO to efficiently explore design spaces while balancing complex tradeoffs between design specifications.
Second, we discuss TSBO, a method that enhances BO’s data efficiency in high-dimensional settings by strategically incorporating unlabeled samples and generating reliable pseudo-labels via a teacher-student model with feedback. TSBO demonstrates strong performance across multiple high-dimensional BO tasks, achieving up to a 364.2x reduction in labeled data usage.
Finally, I will conclude with a discussion on the broader implications and future directions for data-efficient BO.
Biography
Yu Wang is a Ph.D. candidate in Computer Engineering at the University of California, Santa Barbara, advised by Professor Peng Li. He received his M.S. from Texas A&M University in 2019 and his B.S. from Fudan University.
Yu’s research focuses on the intersection of advanced machine learning and hardware system design, with an emphasis on Bayesian Optimization, Semi-Supervised Learning, and hallucination mitigation in Large Language Models (LLMs). He explores their applications in electronic design automation (EDA) to enhance design efficiency and automation.
His works have been recognized at leading machine learning and circuit design venues, including AAAI, ICML, TMLR, and ICCAD. He also received the Best Paper Award at ASAP 2020.
Please join us on Friday, 02/21/25 at 10:20 a.m. in ETB 1020 to learn more on the research presented by CESG’s former M.S. ECE graduate Yu Wang!
Host: Alex Sprintson; Faculty may request a Zoom Link for this presentation.
Wang’s Google Scholar: https://scholar.google.com/citations?user=lUd8s0QAAAAJ&hl=en