Overview
Explore the concept of Residual Based Sampling for online low rank approximation in this 23-minute IEEE conference talk. Delve into topics such as PCA, low rank approximation, column subset selection, and online algorithms. Learn about the motivation behind online clustering and understand the framework presented for this approach. Discover the main idea of accumulating vectors with p 1 and how to summarize vectors effectively. Examine the process of bounding errors and consider open questions in the field. Gain insights into how all these elements come together to form a comprehensive understanding of this sampling technique for online low rank approximation.
Syllabus
Intro
PCA/ Low rank approximation
Column subset selection (CSS)
Online algorithms
What is known?
Motivation online clustering Myerson 2001
Setting for the sake of talk
Framework
Main idea: accumulate vectors with p 1
Summary of vectors
Bounding the error
Open questions
Putting everything together
Taught by
IEEE FOCS: Foundations of Computer Science