ČeskyEnglish

Úvod » Semináre » ACM TechTalks: Machine Learning for Algorithm Design

ACM TechTalks: Machine Learning for Algorithm Design

Kedy: 26. 10. 2021, 18:00 – 19:00
Kde: online prostredníctvom služby On24
Vložné: zdarma, nutná registrácia

Zaregistrujte sa na webinár.

Otázky k webináru ACM TechTalks môžete zadať vopred cez ACM Practitioners Board.

Webináre zo série ACM TechTalks sú určené študentom a odborníkom na počítačovú vedu a používateľom ACM Digital Library. Sú v angličtine online priamo vo vašom internetovom prehliadači a na prenos zvuku stačí, aby ste mali k počítaču pripojené reproduktory alebo slúchadlá.

Nevyhovuje vám čas? Pozrite si webináre TechTalks zo záznamu.

Máte otázky k ACM Digital Library? Dohodnite si osobné/online stretnutie alebo seminár pre vašu inštitúciu.

Invitation

The classic textbook approach to designing and analyzing algorithms assumes worst-case instances of the problem, about which the algorithm designer has absolutely no information at all.

Unfortunately, for many problems such worst-case guarantees—either for solution quality or running time or other performance measures—are often weak. Consequently, rather than using off the shelf algorithms that have weak worst-case guarantees, practitioners often employ a data-driven algorithm design approach; specifically, given an application, they use machine learning and instances of the problem from the specific domain to learn a method that works best in that domain. Historically, such algorithmic techniques have come with no performance guarantees.

In this talk, I will describe our recent work that helps put data-driven algorithm design on firm foundations. I will describe both specific case studies and general principles applicable broadly to a variety of combinatorial problems.

  • Speaker: Maria Florina Balcan, Cadence Design Systems Professor of Computer Science, Carnegie Mellon University
  • Moderator: Steve Hanneke, Assistant Professor, Purdue University