Explore a 49-minute lecture on learning-based algorithms for low-rank approximation problems presented by Piotr Indyk from the Massachusetts Institute of Technology. Delve into the recent advancements in optimizing algorithm performance using training sets of input matrices. Discover the two-step process of efficient approximate algorithms for computing low-rank approximations, involving the computation of a "sketch" and subsequent singular value decomposition. Learn how replacing random matrices with "learned" matrices significantly reduces approximation errors. Gain insights into joint work with Peter Bartlett, Yang Yuan, Ali Vakilian, Tal Wagner, and David Woodruff in this IPAM workshop on Multi-Modal Imaging with Deep Learning and Modeling.
Learning-Based Low-Rank Approximations - IPAM at UCLA
Institute for Pure & Applied Mathematics (IPAM) via YouTube
Overview
Syllabus
Piotr Indyk - Learning-Based Low-Rank Approximations - IPAM at UCLA
Taught by
Institute for Pure & Applied Mathematics (IPAM)
Related Courses
-
Robust and Sample Optimal Algorithms for PSD Low-Rank Approximation
-
Foundations of Data Science II
-
LoRA - Low Rank Adaptation for Parameter Efficient Fine-Tuning
-
Parameter-Efficient Fine-Tuning with LoRA - Optimizing LLMs for Local GPU Training
-
Sampling-Based Sublinear Low-Rank Matrix Arithmetic Framework for Dequantizing Quantum Machine Learning
