Best Arm Identification in Graphical Bilinear Bandits

Authors: Geovani Rizk, Albert Thomas, Igor Colin, Rida Laraki, Yann Chevaleyre

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

Reproducibility Variable Result LLM Response
Research Type Experimental We propose a decentralized allocation strategy based on random sampling with theoretical guarantees. In particular, we characterize the influence of the graph structure (e.g. star, complete or circle) on the convergence rate and propose empirical experiments that confirm this dependency. Finally, Section 7 evidences the theoretical findings on numerical experiments.
Researcher Affiliation Collaboration 1PSL Universit e Paris Dauphine, CNRS, LAMSADE, Paris, France 2Huawei Noah s Ark Lab 3Liverpool University. Correspondence to: Geovani Rizk <geovani.rizk@dauphine.psl.eu>.
Pseudocode Yes Algorithm 1 Bipartite graph algorithm for Best Arm Identification in Graphical Bilinear Bandits Algorithm 2 Randomized G-Allocation strategy for Graphical Bilinear Bandits
Open Source Code No The paper does not provide any concrete access to source code (e.g., specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described.
Open Datasets No The paper describes the creation of the experimental setup and the parameters used (e.g., node-arms, parameter matrix M, noise distribution) but does not refer to a publicly available dataset with concrete access information (link, DOI, 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) for training, validation, or testing.
Hardware Specification No The paper does not provide specific hardware details (e.g., 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 We consider d + 1 node-arms in X Rd where d 2. This node-arm set is made of the d vectors (e1, . . . , ed) forming the canonical basis of Rd and one additional arm xd+1 = (cos(ω), sin(ω), 0, . . . , 0) with ω ]0, π/2]. The parameter matrix M has its first coordinate equal to 2 and the others equal to 0 which makes θ = vec (M ) = (2, 0, . . . , 0) Rd2. We set η(i,j) t N(0, 1), for all edges (i, j) and round t. We consider the two cases where ω = 0.1 and ω = π/2.