数学科学研究所
Insitute of Mathematical Science

Seminar: Convergence and superconvergence of Random Feature Methods

Seminar| Institute of Mathematical Sciences

Time: Thursday, December 18th, 2025,16:00-17:00

LocationRS408, IMS

Speaker: Pingbing Ming, AMSS at CAS

  

Abstract: Among the various machine learning methods solving partial differential equations, the Random Feature Method (RFM) stands out due to its accuracy and efficiency. In this talk, we demonstrate that the approximation error of RFM exhibits spectral convergence when it is applied to the second-order elliptic equations in one dimension, provided that the solution belongs to Gevrey classes or Sobolev spaces. We highlight the significant impact of incorporating the Partition of Unity Method (PUM) to enhance the convergence of RFM by establishing the convergence rate in terms of the maximum patch size. Furthermore, we reveal that the singular values of the random feature matrix (RFMtx) decay exponentially, while its condition number increases exponentially as the number of the features grows. We also theoretically illustrate that PUM may mitigate the excessive decay of the singular values of RFMtx. We shall also discuss the extension of these results to high dimension. This is a joint work with Hao Yu.

地址:上海市浦东新区华夏中路393号
邮编:201210
上海市徐汇区岳阳路319号8号楼
200031(岳阳路校区)