Streaming and randomized techniques for low-rank approximation of matrices and tensors with applications

• Date: 20. 3. 2025
• Time: 14:00 - 15:30
• Address:
  Sokolovská 83, Praha
• Room: K1
• Speaker: Alberto Bucci

In this talk, we will review popular techniques for the low-rank approximation of matrices and tensors and explore their extension to a streamable setting, where data can only be accessed once, requiring efficient, single-pass algorithms. We will provide theoretical insights and performance analysis. We will then focus on a recent research direction where these streamable techniques play a crucial role: solving linear systems and least squares problems in various low-rank formats. In particular, we will discuss low-rank sketched GMRES and sketched LSQR, highlighting their advantages and challenges and their connection with the recent developed star process.