Stochastic Rising Bandits

Authors: Alberto Maria Metelli, Francesco Trovò, Matteo Pirola, Marcello Restelli

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically compare our algorithms with state-of-the-art methods for non-stationary MABs over several synthetically generated tasks and an online model selection problem for a realworld dataset.
Researcher Affiliation Academia 1Dipartimento di Elettronica, Informazione e Bioingegneria, Politecnico di Milano, Milan, Italy.
Pseudocode Yes Algorithm 1 R-l-UCB (l Ptless,edu )
Open Source Code Yes The code to reproduce the experiments is available at https://github.com/albertometelli/ stochastic-rising-bandits.
Open Datasets Yes We employ the IMDB dataset, made of 50,000 reviews of movies (scores from 0 to 10). We preprocessed the data as done by Maas et al. (2011) to obtain a binary classification problem.
Dataset Splits No The paper describes the use of datasets for evaluation, but it does not specify explicit training, validation, and test splits (e.g., percentages or counts) or reference a standard splitting methodology.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments (e.g., GPU models, CPU specifications, or memory).
Software Dependencies No The paper mentions various algorithms and packages (e.g., 'KL-UCB', 'Ser4', 'SW-UCB', 'SW-TS') but does not specify the version numbers for any software dependencies required to reproduce the experiments.
Experiment Setup Yes The parameters for all the baseline algorithms have been set as recommended in the corresponding papers (see also Appendix E). For our algorithms, the window is set as hi,t tϵNi,t 1u (as prescribed by Theorems 4.4 and 5.3). ...setting ϵ 1{4. ...R-ed-UCB, with ϵ 1{32.