报告题目：Continuous Normalizing Flow for Conditional  Generative Learning

报告时间：2025-03-06   16:30-17:30

报  告 人 ：黄坚   教授（香港理工大学）

报告地点：雷军科技楼六楼报告厅（644）

Abstract：Continuous Normalizing Flows (CNFs) are a generative modeling technique that utilizes ordinary differential equations to learn probability distributions. This approach has been successful in a range of applications, including image synthesis, protein structure prediction, and molecule generation. In this talk, we will present the CNF method and explore its theoretical properties through a flow matching objective function. We will then introduce a conditional CNF method and demonstrate its application in controlled image generation by fine-tuning Stable Diffusion 3, a large foundational image model.