Categorized Bandits

Authors: Matthieu Jedor, Vianney Perchet, Jonathan Louedec

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we have conducted an analysis on real data to highlight that those ordered categories actually exist in practice. Finally, we conduct numerical experiments on different scenarios to illustrate both finite-time and asymptotic performances of our algorithms compared to algorithms either agnostic to the structure or only taking it partly into account.
Researcher Affiliation Collaboration Matthieu Jedor CMLA, ENS Paris-Saclay & Cdiscount jedor@cmla.ens-cachan.fr Jonathan Louëdec Cdiscount jonathan.louedec@cdiscount.com Vianney Perchet CMLA, ENS Paris-Saclay & Criteo AI Lab perchet@cmla.ens-cachan.fr
Pseudocode Yes Algorithm 1: CATSE(δ)
Open Source Code No The paper does not provide concrete access to source code for the described methodology. No links to repositories or explicit statements of code release are found.
Open Datasets No We have collected the CTR of products in four different categories over one month on the e-commerce website Cdiscount, one of the leading ecommerce companies in France, gathered in Table 2a. For privacy reason, the exact content of the different categories cannot be revealed.
Dataset Splits No The paper describes numerical experiments where rewards are drawn from Gaussian distributions or analysis of collected real-world data (for which privacy is cited as a reason for non-disclosure), but does not specify any training/validation/test dataset splits.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. It only states that simulations were run.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names with versions) needed to replicate the experiment.
Experiment Setup Yes The simulations were ran until time horizon 10,000 and results were averaged over 100 independent runs. CATSEs and CATSE0 were run with δ = 1/t and δ = 1/Mt, respectively.