MDS20 Minitutorial: A Mathematical Perspective of Machine Learning by Weinan E
Abstract: The heart of modern machine learning is the approximation of high dimensional functions. Traditional approaches, such as approximation by piecewise polynomials, wavelets, or other linear combinations of fixed basis functions, suffer from the curse of dimensionality. We will discuss representations and approximations that overcome this difficulty, as well as gradient flows that can be used to find the optimal approximation. We will see that at the continuous level, machine learning consists of a series of reasonably nice variational and PDE-like problems. Modern machine learning models/algorithms, such as the random feature and shallow/deep neural network models, are all special discretizations of these continuous problems. We will also discuss how to construct new models/algorithms using the same philosophy. At the theoretical level, we will present a framework that is suited for analyzing machine learning models and algorithms in high dimension, and present results that are free of the curse of dimensionality. Finally, we will discuss the fundamental reasons that are responsible for the success of modern machine learning, as well as the subtleties and mysteries that still remain to be understood.
Weinan E, Princeton University, U.S., weinan@math.princeton.edu
This is one of six minitutorial talks organized by Carola-Bibiane Schönlieb (University of Cambridge, United Kingdom) and Ozan Öktem (KTH Stockholm, Sweden) under the title "Deep Learning for Inverse Problems and Partial Differential Equations" as part of the 2020 SIAM Conference on Mathematics of Data Science. For more information, visit https://meetings.siam.org/sess/dsp_programsess.cfm?SESSIONCODE=68215.
Видео MDS20 Minitutorial: A Mathematical Perspective of Machine Learning by Weinan E канала SIAM Conferences
Weinan E, Princeton University, U.S., weinan@math.princeton.edu
This is one of six minitutorial talks organized by Carola-Bibiane Schönlieb (University of Cambridge, United Kingdom) and Ozan Öktem (KTH Stockholm, Sweden) under the title "Deep Learning for Inverse Problems and Partial Differential Equations" as part of the 2020 SIAM Conference on Mathematics of Data Science. For more information, visit https://meetings.siam.org/sess/dsp_programsess.cfm?SESSIONCODE=68215.
Видео MDS20 Minitutorial: A Mathematical Perspective of Machine Learning by Weinan E канала SIAM Conferences
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
IS20: IP5: “Deep Internal Learning” -- Deep Learning with Zero ExamplesIS20: IP4: Feathers, Wings, and the Future of Computer VisionMDS20 – Growing Graphs and Topological Robustness to Node Order PerturbationAN20: Multi-Applications of Multi-OrthogonalitySIAM Conference on Mathematics of Planet Earth (MPE20) Invited Plenary Talks: Alan M. HastingsSIAM Conference on Life Sciences (LS20) Invited Plenary Talks: Jeremy GunawardenaIS20: IP6: Deep Learning in Wave-based Imaging and Inverse ProblemsIS20: MT3: Data Driven Methods for MRI ReconstructionAN20: Elements of An Effective Ecosystem of Math Libraries for ExascalePP20 - Nur A Fadel - Utopia: a Performance-Portable C++ Library for Non-Linear AlgebraCSE19 - Boyce GriffithSIAM Conference on Mathematics of Planet Earth (MPE20) Invited Plenary Talks: Elizabeth HunkeAN20: Theodore von Kármán Prize Lecture: Mathematics, Mechanics and MaterialsAN20: Structure and Scale Interactions in Wall Turbulence: A Modal PuzzleSIAM Mathematics of Data Science (MDS20) Distinguished Lecture Series: Jennifer Tour ChayesPP20 - Mawussi Zounon - Opportunities for Multi Precision Computation in Memory Bound ApplicationsSIAM Mathematics of Data Science (MDS20) Distinguished Lecture Series: Yann LeCunPP20 - Sarah Knepper - Intel Software Solutions for Diverse Computing ArchitecturesIS20: IP2: Variational NetworksSIAM Mathematics of Data Science (MDS20) Distinguished Lecture Series: David L. DonohoI. E. Block Lecture: Erik Demaine