This is an online seminar. Registration is required.
【Deep Learning Theory Team】
【Date】2025/May 9 (Fri) 18:00-19:00(JST)
【Speaker】Leyang Wang, University College London, Computer Science Department,Master student
Title: Differential Parameter Inference in Exponential Family using Time Score Matching
Abstract:
This work addresses differential inference in time-varying parametric probabilistic models, like graphical models with changing structures. Instead of estimating a high-dimensional model at each time and inferring changes later, we directly learn the differential parameter, i.e., the time derivative of the parameter. The main idea is treating the time score function of an exponential family model as a linear model of the differential parameter for direct estimation. We use time score matching to estimate parameter derivatives. We prove the consistency of a regularized score matching objective and demonstrate the finite-sample normality of a debiased estimator in high-dimensional settings. Two applications will be presented: one on learning differential graphical models and the other on guiding generative models with natural gradients. If time permits, an ongoing work on diffusion distillation using a variational approach will also be presented.
