Timezone: »

Anytime Model Selection in Linear Bandits
Parnian Kassraie · Nicolas Emmenegger · Andreas Krause · Aldo Pacchiano

Wed Dec 13 03:00 PM -- 05:00 PM (PST) @ Great Hall & Hall B1+B2 #924
Model selection in the context of bandit optimization is a challenging problem, as it requires balancing exploration and exploitation not only for action selection, but also for model selection. One natural approach is to rely on online learning algorithms that treat different models as experts. Existing methods, however, scale poorly ($\mathrm{poly}M$) with the number of models $M$ in terms of their regret.Our key insight is that, for model selection in linear bandits, we can emulate full-information feedback to the online learner with a favorable bias-variance trade-off. This allows us to develop ALEXP, which has an exponentially improved ($\log M$) dependence on $M$ for its regret.ALEXP has anytime guarantees on its regret, and neither requires knowledge of the horizon $n$, nor relies on an initial purely exploratory stage.Our approach utilizes a novel time-uniform analysis of the Lasso, establishing a new connection between online learning and high-dimensional statistics.

Author Information

Parnian Kassraie (ETH Zurich)
Parnian Kassraie

I am a Ph.D. student at the Department of Computer Science at ETH Zurich, advised by [Andreas Krause](https://las.inf.ethz.ch/krausea). I am part of the [Institute for Machine Learning](https://ml.inf.ethz.ch/) and associated with the [ETH AI Center](https://ai.ethz.ch/). The primary focus of my research is in the foundations of Reinforcement Learning. I joined ETH as a Direct Doctorate student and in Spring 2021 completed my M.Sc. under the supervision of Andreas Krause and [Fanny Yang](http://fanny-yang.de/), with a thesis on Contextual Neural Bandits. Before ETH, I studied at the Sharif University of Technology where I earned a dual B.Sc. in Electrical Engineering and Computer Science.

Nicolas Emmenegger (ETH Zurich)
Andreas Krause (ETH Zurich)
Aldo Pacchiano (Boston University / Broad Institute of MIT and Harvard)

More from the Same Authors