Games for Computation and Learning
Abstract
This proposal is to further develop a game theoretic approach to numerical approximation and algorithm design and apply it to the design of machine learning algorithms.Prior work on this game theoretic approach has lead to the discoveries of (i) wavelets adaptedto arbitrary linear operators (gamblets) (ii) scalable solvers with some degree of universality (iii) new tools for numerical analysis and algorithm design such as the Fast Gamblet Transform (FGT).
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 28, 2018
- Source ID
- FA95501810271
Entities
People
- Houman Owhadi
Organizations
- Air Force Office of Scientific Research
- California Institute of Technology
- United States Air Force