Scalable Discrete Sampling as a Multi-Armed Bandit Problem

Authors: Yutian Chen, Zoubin Ghahramani

ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical evaluations show the robustness and efficiency of the approximate algorithms in both synthetic and real-world large-scale problems.5. Experiments
Researcher Affiliation Academia Yutian Chen YUTIAN.CHEN@ENG.CAM.AC.UK Department of Engineering, University of Cambridge, Cambridge CB2 1PZ, UK Zoubin Ghahramani ZOUBIN@ENG.CAM.AC.UK Department of Engineering, University of Cambridge, Cambridge CB2 1PZ, UK Alan Turing Institute, 96 Euston Road, London NW1 2DB, UK
Pseudocode Yes Alg. 1 Racing Algorithm with a Finite Reward Population
Open Source Code No B provides a lookup table and a plot of BNormal(δ) = E 1(δ). Notice that BNormal only needs to be computed once and we can obtain it for any δ by either interpolating the table or computing numerically with code to be shared (runtime < 1 second).
Open Datasets No The paper mentions 'DBLP dataset' and 'Rexa corpus' but does not provide specific access information (link, DOI, repository, or formal citation for the dataset itself).
Dataset Splits No The paper does not provide specific training/test/validation dataset splits (e.g., percentages, sample counts, or explicit cross-validation setup).
Hardware Specification No No specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, memory amounts) used for running experiments are mentioned.
Software Dependencies No No specific ancillary software details (e.g., library or solver names with version numbers) are provided.
Experiment Setup Yes We set m(1) = 50 and δ = 0.05. We use adjusted priors q as suggested by Carlin & Chib (1995) for sufficient mixing between all models and tune them with adaptive MCMC.