Uploaded March 2024 | Updated September 2026, 1 week ago
Optimization has proved to be a rich source of techniques for formulating and solving computational problems that arise in data analysis and machine learning. This tutorial surveys problems in kernel learning, regression, graph analysis, neural networks, low-rank matrix analysis, and other areas that can be formulated as optimization problems over spaces of real vectors or matrices. We touch on the role of regularization in promoting useful solution structures. Finally, we describe the primary algorithmic techniques, focusing gradient and stochastic gradient methods.
MT1: Optimization for Data Analysis
Organizer: Stephen Wright
University of Wisconsin-Madison, U.S.
This talk was given at the 2022 SIAM Conference on Mathematics of Data Science in San Diego, California, U.S. Learn more about SIAM Conferences at siam.org/conferences/about-siam-conferences
0:00 Part 1 of Minitutorial
55:36 Part 1 Question and Answers
56:54 Part 2 of Minitutorial
Optimization has proved to be a rich source of techniques for formulating and solving computational problems that arise in data analysis and machine learning. This tutorial surveys problems in kernel learning, regression, graph analysis, neural networks, low-rank matrix analysis, and other areas that can be formulated as optimization problems over spaces of real vectors or matrices. We touch on the role of regularization in promoting useful solution structures. Finally, we describe the primary algorithmic techniques, focusing gradient and stochastic gradient methods.
MT1: Optimization for Data Analysis
Organizer: Stephen Wright
University of Wisconsin-Madison, U.S.
This talk was given at the 2022 SIAM Conference on Mathematics of Data Science in San Diego, California, U.S. Learn more about SIAM Conferences at siam.org/conferences/about-siam-conferences
0:00 Part 1 of Minitutorial
55:36 Part 1 Question and Answers
56:54 Part 2 of Minitutorial
