An Algorithm for Stochastic and Adversarial Bandits with Switching Costs

Authors: Chloé Rouyer, Yevgeny Seldin, Nicolò Cesa-Bianchi

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We provide experimental evaluation showing competitiveness of our algorithm with the relevant baselines in the stochastic, stochastically constrained adversarial, and adversarial regimes with fixed switching cost. 7. Experiments We compare the performance of Tsallis-Switch to different baselines, both in the stochastic and in the stochastically constrained adversarial regime.
Researcher Affiliation Academia 1Department of Computer Science, University of Copenhagen, Denmark 2DSRC & Dept. of Computer Science, Universit a degli Studi di Milano, Milano, Italy.
Pseudocode Yes Algorithm 1 Tsallis-Switch Input: Learning rates η1 η2 > 0. Block lengths |B1|, |B2|, . . . . Initialize: C0 = 0K for n = 1, 2, . . . do pn = arg min p K 1 Sample In pn and play it for all rounds t Bn. Observe and suffer cn,In = P t Bn ℓt,In. i [K] : cn,i = pn,i , if In = i, 0, otherwise. i [K] : Cn(i) = Cn 1(i) + cn,i. end for
Open Source Code No The information is insufficient as the paper does not contain any explicit statement about providing open-source code for the described methodology, nor does it include a link to a code repository.
Open Datasets No The information is insufficient. The paper describes generating synthetic data (“We generate binary losses using two sets of parameters...”) but does not provide access information (link, DOI, formal citation) for a publicly available dataset.
Dataset Splits No The information is insufficient as the paper does not specify exact dataset split percentages, absolute sample counts, or reference predefined splits for training, validation, or testing.
Hardware Specification No The information is insufficient as the paper does not specify any particular hardware components (e.g., GPU models, CPU models, or cloud resources with specifications) used for running the experiments.
Software Dependencies No The information is insufficient as the paper does not provide specific software dependencies with version numbers (e.g., library names like PyTorch or TensorFlow, along with their versions).
Experiment Setup No The information is insufficient as the paper describes the characteristics of the problem instances used in experiments (e.g., “number of arms K = 8”, “gaps = 0.2 and λ = 0.025”) but does not provide specific experimental setup details such as concrete hyperparameter values or system-level training settings.