报告时间：2024年4月23日（周二）14:00-15:00
报告地点：中国人民大学明德主楼1037
报告主题：Conditional Stochastic Interpolation for Generative Learning
报告摘要：
We propose a conditional stochastic interpolation (CSI) approach for learning conditional distributions. CSI learns probability flow equations or stochastic differential equations that transport a reference distribution to the target conditional distribution. This is achieved by first learning the velocity function and the conditional score function based on conditional stochastic interpolation, which are then used to construct a deterministic process governed by an ordinary differential equation or a diffusion process for conditional sampling. We establish the transport equation and derive the explicit form of the conditional score function with mild conditions. We also incorporate an adaptive diffusion term in our proposed CSI model to address the instability issues arising during the training process. Furthermore, we establish non-asymptotic error bounds for learning the target conditional distribution via conditional stochastic interpolation in terms of KL divergence, taking into account the neural network approximation error. We illustrate the application of CSI on image generation using benchmark image data.
报告人简介：
Dr. Li is an assistant professor in the Department of Applied Mathematics at Hong Kong Polytechnic University. Prior to joining PolyU, he was a postdoctoral associate in Yale University, Biostatistics Department. He received his PhD in Hong Kong University of Science and Technology. His research focuses on data science and statistical learning on complex data, especially on network data, brain data and imaging genomics, generative learning and large models.
