Id: arxiv/2606.26652 Url: https://arxiv.org/abs/2606.26652 Tags: math.ST, stat.ML Score: 0 Title: Scalable Operator Learning via Nyström Approximation With Denoising Applications Domain: math.ST Source: arxiv Authors: Naveen Gupta, Vaibhav Silmana, S. Sivananthan Reviews: 0 Subject: math.ST Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-06-25 Page: https://delphi.science/r/arxiv/2606.26652 In this paper, we study Nyström subsampling for vector-valued regression in vector-valued reproducing kernel Hilbert spaces. Standard kernel methods often suffer from prohibitive computational costs due to the construction and inversion of large kernel matrices, which limits their scalability to large datasets. To overcome this bottleneck, we propose an efficient operator learning algorithm based on Nyström subsampling that accommodates functional outputs. Under general source conditions characterized by index functions-extending beyond the classical Hölder-type and operator-monotone…