log in  |  register  |  feedback?  |  help  |  web accessibility
The Many Faces of Optimal Weak-to-Strong Learning
Michael Xie - University of Maryland, College Park
IRB 3137 or Zoom: https://umd.zoom.us/j/6778156199?pwd=NkJKZG1Ib2Jxbmd5ZzNrVVlNMm91QT09
Thursday, December 5, 2024, 2:30-3:30 pm
  • You are subscribed to this talk through .
  • You are watching this talk through .
  • You are subscribed to this talk. (unsubscribe, watch)
  • You are watching this talk. (unwatch, subscribe)
  • You are not subscribed to this talk. (watch, subscribe)
Abstract

In this presentation we will study the concept of weak-to-strong learners, with an emphasis on boosting. In particular, we will focus on a paper by Mikael Møller Høgsgaard, Kasper Green Larsen, Markus Engelund Mathiasen which presents a surprisingly simple boosting algorithm that achieves optimal sample complexity. Previous work has shown a theoretical lower bound for sample complexity. Sample optimal boosting algorithms have only recently been developed, but we will show that this new algorithm has faster runtime while being simpler to implement.

Bio

Michael is a first year MS student at the University of Maryland advised by Aravind Srinivasan, focusing on algorithm design and analysis. He is broadly interested in randomized algorithms and probabilistic methods.

This talk is organized by Kishen N Gowda