报告题目 Title:Analysis and Control of Diffusion Models: From Pretraining to Post-Training and Posterior Sampling
报告人 Speaker:庞彤瑶
报告人所在单位 Affiliation:清华大学
时间 Time:2026-11-25 14:30-15:30
地点 Venue:腾讯会议:159-307-655 会议密码:261114
报告摘要 Abstract:This talk studies diffusion models across the pipeline from pretraining to post-training and posterior sampling. We first analyze how the forward diffusion schedule affects generation accuracy and sampling complexity, providing guidance for the design of stronger base models. Given a pretrained model, we then consider two ways to steer its generative distribution toward specific objectives: reinforcement-learning post-training for reward alignment and direct posterior sampling for inverse problems. Together, these topics illustrate how diffusion models can be designed, analyzed, and controlled from base-model construction to downstream applications.
个人简介 Bio:庞彤瑶,现为清华大学丘成桐数学科学中心助理教授。2014年本科毕业于北京大学,2019年获新加坡国立大学数学博士学位。主要研究方向为深度生成模型的理论、算法及其在图像处理与反问题中的应用,重点关注扩散模型的采样与误差分析、强化学习后训练以及后验采样。
海报 Poster:
庞彤瑶 学术报告.jpg
电话 Tel:021-65648958
邮箱 Email:am_admin@fudan.edu.cn
地址 Address:上海市杨浦区湾谷科技园二期D1栋
Building D1, Bay Valley II, Yangpu District, Shanghai, China
邮编 Postcode:200438