Bandits with many optimal arms

Authors: Rianne de Heide, James Cheshire, Pierre Ménard, Alexandra Carpentier

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Numerical Experiments We compare our algorithm against other standard bandit algorithms via numerical simulations. The results reported are averaged over 100 independent trials.
Researcher Affiliation Collaboration Alon Cohen: Massachusetts Institute of Technology, Cambridge, MA, USA, alonco@mit.edu. Yishay Mansour: Google and Tel Aviv University, Tel Aviv, Israel, mansour.yishay@gmail.com.
Pseudocode No The paper describes the algorithm mathematically and textually, but does not include a dedicated 'Pseudocode' or 'Algorithm' block with structured steps.
Open Source Code No The paper does not contain any statement about releasing source code or provide a link to a code repository for the described methodology.
Open Datasets No The paper conducts numerical simulations rather than using a publicly available dataset. It describes the simulation setup (e.g., '100 independent trials', 'Gaussian distribution with mean µ and variance σ2 = 0.01'). There is no external dataset to provide access information for.
Dataset Splits No The paper discusses numerical simulations and hyperparameter tuning but does not explicitly provide details on train/validation/test dataset splits, as is common in supervised learning. It mentions: 'The hyperparameter C was selected to achieve the best empirical performance on the range of parameters we tested.'
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to run the numerical experiments.
Software Dependencies No The paper does not specify any software dependencies with version numbers (e.g., programming languages, libraries, or solvers).
Experiment Setup Yes The paper provides details on the experimental setup, including the number of trials ('averaged over 100 independent trials'), the horizon ('T = 10000'), number of arms ('K = 100'), and the distribution parameters for rewards ('Gaussian distribution with mean µ and variance σ2 = 0.01'). It also mentions that 'The hyperparameter C was selected to achieve the best empirical performance on the range of parameters we tested.'