Non-Stationary Delayed Bandits with Intermediate Observations

Authors: Claire Vernade, Andras Gyorgy, Timothy Mann

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results demonstrate that our method is able to learn in non-stationary delayed environments where existing methods fail.
Researcher Affiliation Industry 1DeepMind, London, UK.
Pseudocode Yes Algorithm 1 NSD-UCRL2 for NSD-Bandits
Open Source Code No The paper does not provide concrete access to source code (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described in this paper.
Open Datasets No The experiments use a synthetic setup based on parameters in Table 1, not a publicly available dataset with concrete access information (link, DOI, repository, or formal citation).
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning. It mentions '50 independent runs' but no train/validation/test splits.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment.
Experiment Setup Yes Throughout the experiments we use the transition probabilities and mean rewards as reported in Table 1. The delay parameter D and the window size W of the algorithm are discussed in the dedicated sections. ... we run experiments with time horizon T = 8000 with ΓT = 3 change points at rounds {2000, 4000, 6000}. In the next experiments, we will set W = 800, which is close to the value suggested by theory and that works best in this calibration experiment.