Covariance-adaptive best arm identification
Authors: El Mehdi Saad, Gilles Blanchard, Nicolas Verzelen
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We introduce new algorithms that adapt to the unknown covariance of the arms and demonstrate through theoretical guarantees that substantial improvement can be achieved over the standard setting. Additionally, we provide new lower bounds for the relaxed setting and present numerical simulations that support their theoretical findings. |
| Researcher Affiliation | Academia | El Mehdi Saad Université Paris-Sacalay, Centrale Supéléc Laboratoire des Signaux et Systèmes Paris, France el-mehdi.saad@centralesupelec.fr Gilles Blanchard Institut de Mathématique d Orsay Université Paris-Saclay Paris, France gilles.blanchard@universite-paris-saclay.fr Nicolas Verzelen Mistea, INRAE Institut Agro, Université de Montpellier Montpellier, France nicolas.verzelen@inrae.fr |
| Pseudocode | Yes | Protocol 1 The Game Protocol; Algorithm 2 Pairwise-BAI |
| Open Source Code | No | The paper does not include an unambiguous statement about releasing code for the described methodology or a link to a source-code repository. |
| Open Datasets | No | We conduct numerical experiments using synthetic data to assess the practical relevance of our approach. |
| Dataset Splits | No | The paper uses synthetic data generated for simulations and does not specify standard train/validation/test splits as one would for fixed supervised learning datasets. No specific percentages, counts, or predefined split citations are provided. |
| Hardware Specification | No | The paper does not specify any hardware details (e.g., CPU, GPU models, memory, or cloud instances) used for running the numerical simulations. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., programming languages, libraries, or frameworks) needed to replicate the experiments. |
| Experiment Setup | Yes | We consider the Gaussian rewards scenario. ... for K = 10 arms with means µi = i/10. The covariance matrix C is defined as follows: for i JKK, Cii = 1 and for i = j: Cij = ρ. We consider 4 scenarios with the correlation: ρ {0, 0.5, 0.7, 0.9}. ... with K = 16 arms with means µi = i/10. Arms in the same cluster have a correlation of 0.99, and arms from different clusters are independent. We consider 4 scenarios with different number of clusters: ncl {8, 4, 2, 1}. |