idw – Informationsdienst Wissenschaft

Nachrichten, Termine, Experten

Grafik: idw-Logo
Grafik: idw-Logo

idw - Informationsdienst
Wissenschaft

Science Video Project
idw-Abo

idw-News App:

AppStore

Google Play Store

Veranstaltung


institutionlogo

04.07.2023 - 04.07.2023 | Saarbrücken

DLS Talk: Smooth Contextual Bandits (Nathan Kallus), July 4th, 10am-12pm

We are very proud to announce that Professor Nathan Kallus (Cornell University) is our guest for the next edition of our “Distinguished Lecture Series”. Prof Kallus will give a talk on "Smooth Contextual Bandits" on Tuesday, 4th of July, at 10am. Contextual bandit problems model the inherent cost of learning in personalized decision-making in new environments, whether in marketing, healthcare, or revenue management.

Abstract:

Contextual bandit problems model the inherent cost of learning in personalized
decision-making in new environments, whether in marketing, healthcare, or revenue
management. Specifically, the cost is characterized by the optimal growth rate of the
regret in cumulative rewards compared to an optimal policy given full prior knowledge
of the environment. Naturally, the optimal rate should depend on how complex the
underlying supervised learning problem is, namely how much can observing rewards in
one context tell us about mean rewards in another context. Curiously, this
obvious-seeming relationship is obscured in current theory that separately studies the
easy, fully-extrapolatable case and hard, super-local case. To characterize the
relationship more precisely, I study a nonparametric contextual bandit problem where
expected reward functions are β-smooth (roughly meaning β-times differentiable). I
will show how this interpolates between the two extremes previously studied in
isolation: non-differentiable-response bandits (β ≤ 1), where rate-optimal regret is
achieved by decomposing the problem into non-contextual bandits, and
parametric-response bandits (β = ∞), where rate-optimal regret is often achievable
without any exploration at all. We develop a novel algorithm that works for any given
smoothness setting by operating neither fully locally nor fully globally. We prove its
regret is rate-optimal, thereby characterizing the optimal regret rate and revealing a
fuller picture of the crucial interplay between complexity and regret in dynamic
decision-making. Time permitting, I will also discuss how to construct valid
confidence intervals from data collected by contextual bandits, a crucial challenge in
the enterprise to replace randomized trials with adaptive experiments in applied
fields from biostatistics to development economics.

This talk is based on the following papers:
(1) Smooth Contextual Bandits: Bridging the Parametric and Non-differentiable Regret
Regimes (https://pubsonline.informs.org/doi/abs/10.1287/opre.2021.2237)
(2) Post-Contextual-Bandit Inference
(https://papers.nips.cc/paper/2021/hash/eff3058117fd4cf4d4c3af12e273a40f-Abstract...)

Short Bio:

Nathan Kallus is an Associate Professor in the School of Operations Research and
Information Engineering and Cornell Tech at Cornell University. He also holds a
Research Director position for the Product Machine Learning Research at Netflix.
Nathan's research interests include personalization; optimization, especially under
uncertainty; causal inference; sequential decision-making; credible and robust
inference; and algorithmic fairness. He holds a PhD in Operations Research from MIT as
well as a BA in Mathematics and a BS in Computer Science both from UC Berkeley. Before
coming to Cornell, Nathan was a Visiting Scholar at USC's Department of Data Sciences
and Operations and a Postdoctoral Associate at MIT's Operations Research and
Statistics group.

Hinweise zur Teilnahme:
The talk will take place in a hybrid mode with a physical presence in the Bernd Therre lecture hall at CISPA’s main building in Saarbrücken (Stuhlsatzenhaus 5) and via Zoom:
https://cispa-de.zoom.us/j/61118095073?pwd=UmM4bnNuamVjeVQwRy9qTDhIbHNyZz09 
ID: 611 1809 5073
Passcode: F8F7%*

Termin:

04.07.2023 10:00 - 12:00

Veranstaltungsort:

CISPA Helmholtz Center for Information Security
Stuhlsatzenhaus 5
Bernd Therre lecture hall (Ground Floor)
66123 Saarbrücken
Saarland
Deutschland

Zielgruppe:

Studierende, Wissenschaftler

Relevanz:

überregional

Sachgebiete:

Informationstechnik

Arten:

Vortrag / Kolloquium / Vorlesung

Eintrag:

29.06.2023

Absender:

Dr. Felix Koltermann

Abteilung:

Unternehmenskommunikation

Veranstaltung ist kostenlos:

ja

Textsprache:

Deutsch

URL dieser Veranstaltung: http://idw-online.de/de/event74735

Anhang
attachment icon DLS Talk: Smooth Contextual Bandits (Nathan Kallus)

Hilfe

Die Suche / Erweiterte Suche im idw-Archiv
Verknüpfungen

Sie können Suchbegriffe mit und, oder und / oder nicht verknüpfen, z. B. Philo nicht logie.

Klammern

Verknüpfungen können Sie mit Klammern voneinander trennen, z. B. (Philo nicht logie) oder (Psycho und logie).

Wortgruppen

Zusammenhängende Worte werden als Wortgruppe gesucht, wenn Sie sie in Anführungsstriche setzen, z. B. „Bundesrepublik Deutschland“.

Auswahlkriterien

Die Erweiterte Suche können Sie auch nutzen, ohne Suchbegriffe einzugeben. Sie orientiert sich dann an den Kriterien, die Sie ausgewählt haben (z. B. nach dem Land oder dem Sachgebiet).

Haben Sie in einer Kategorie kein Kriterium ausgewählt, wird die gesamte Kategorie durchsucht (z.B. alle Sachgebiete oder alle Länder).