Machine Learning via Dynamical Systems

Abstract

The PI proposes to develop a new approach to machine learning via dynamical systems. This can be considered as a generalization of t"he deep neural networks to the continuous setting. But at the same time, it also allows for various extensions and algorithmic impro""vement. Besides offering new possibilities in practical applications, this new approach is very appealing to the applied mathematics"" community and at the same time, it will help to clarify many issues in machine learning. As an application, the PI proposes a very" new idea for solving partial differential equations in highdimensions using the dynamical system-based machine learning approach.The proposed project opens a major new avenue to think about machine learning and other modeling problems that suffer from the curs"e of dimensionality, such as high dimensional partial differential equations, stochastic control problems, filtering, uncertainty qu""antification, etc. It is particularly appealing to the applied mathematics community. It is the PI s hope that this will help to bri"dge the gap between machine learning and traditional applied mathematics.

Document Details

Document Type
DoD Grant Award
Publication Date
Sep 29, 2017
Source ID
N000141712926

Entities

People

  • Weinan E

Organizations

  • Office of Naval Research
  • Trustees of Princeton University
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Neural Network Machine Learning.
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms
  • AI & ML - Neural Networks