Optimal No-Regret Learning for One-Sided Lipschitz Functions

Authors: Paul Duetting, Guru Guruganesh, Jon Schneider, Joshua Ruizhi Wang

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical Our primary contribution is the construction of an efficient algorithm for this problem with an asymptotically optimal regret bound. In particular, we show it is possible to optimize any L-left(/right)-Lipschitz continuous function from [0, 1] to [0, L] while incurring at most O(L log log T) regret: Theorem 1.1 (Restatement of Corollary 3.2 and Theorem 5.1). There exists a deterministic and efficient online algorithm (Algorithm 2) for maximizing an L-left-Lipschitz continuous function mapping [0, 1] [0, L] over T rounds that incurs at most O(L log log T) regret. Moreover, any learning algorithm must incur at least Ω(L log log T) for this problem.
Researcher Affiliation Industry Paul D utting * 1 Guru Guruganesh * 2 Jon Schneider * 3 Joshua R. Wang * 2 *Equal contribution 1Google Research, Z urich, Switzerland 2Google Research, Mountain View, USA 3Google Research, New York City, USA. Correspondence to: Joshua R. Wang <joshuawang@google.com>.
Pseudocode Yes Algorithm 1 Sweep Using Maximum Value Hint, Algorithm 2 Hybrid Binary and KL Search. These are presented in pseudocode format in the paper.
Open Source Code No The paper does not provide any concrete statement about the availability of its source code or a link to a code repository.
Open Datasets No This paper is theoretical and does not conduct experiments involving datasets, therefore no information about public dataset availability is provided.
Dataset Splits No This paper is theoretical and does not conduct experiments, therefore no information about training/test/validation splits is provided.
Hardware Specification No This is a theoretical paper describing algorithms and proofs, and does not involve experimental runs requiring specific hardware. Therefore, no hardware specifications are mentioned.
Software Dependencies No This is a theoretical paper and does not discuss software implementations or dependencies with version numbers for experimental reproducibility.
Experiment Setup No This is a theoretical paper that focuses on algorithm design and proofs, and does not describe an experimental setup with specific details like hyperparameters or training configurations.