How Much Data Is Sufficient to Learn High-Performing Algorithms?
Institute for Pure & Applied Mathematics (IPAM) via YouTube
Overview
Syllabus
Intro
Data-driven algorithm design
Sequence alignment algorithms
Automated configuration
This talk: Main result
Domains with piecewise structure
Primary challenge in combinatorial domains
Example: Sequence alignment
Algorithmic performance
Generalization bounds
Piecewise constant utility function
Primal & dual classes
Warmup: 1-dimensional parameters
Intrinsic complexity
Main result (informal)
Outline
Piecewise constant dual functions
Taught by
Institute for Pure & Applied Mathematics (IPAM)