Statistical and Computational Trade-off in Multi-Agent Multi-Armed Bandits

Authors: Filippo Vannella, Alexandre Proutiere, Jaeseong Jeong

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We devise Efficient Sampling for MAMAB (ESM), an algorithm whose regret asymptotically matches the approximated lower bounds. The regret and computational complexity of ESM are assessed numerically, using both synthetic and real-world experiments in radio communications networks. In this section, we present numerical experiments to assess the performance of our algorithm.
Researcher Affiliation Collaboration Filippo Vannella KTH and Ericsson Research Stockholm, Sweden vannella@kth.se Alexandre Protiuere KTH Stockholm, Sweden alepro@kth.se Jaeseong Jeong Ericsson Research Stockholm, Sweden jaeseong.jeong@ericsson.com
Pseudocode Yes The pseudocode of ESM is presented in Alg. 1.
Open Source Code Yes The code for the synthetic experiments and the additional experiments presented in App. K is available at this link.
Open Datasets No We consider the factor graphs depicted in Fig. 3. The expected rewards are selected uniformly at random in the interval [0, 10]. We run our experiments in a proprietary mobile network simulator in an urban environment. The paper does not provide access to the data generated or used, explicitly stating the network simulator is 'proprietary'.
Dataset Splits No The paper describes generating synthetic data and using a proprietary simulator for experiments, but does not provide details on training, validation, or test dataset splits.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No We implement the solver for the lower bound optimization problems using CVXPY [17] with a MOSEK solver [1]. The versions (e.g., MOSEK 9.0 from bibliography reference [1]) are not directly stated in the main text's experimental setup description.
Experiment Setup Yes In our experiments, select N = 5 and K = 3. We execute our experiments for Nsim = 5 independent runs. Following previous work [8], we select γ = 0, and ε = 0.01. The elimination order is chosen as O = [N]. We test our algorithm for Ai = {2 , 7 , 13 }, and for |S| = 6 sectors. As the factor graph contains cycles, we use = MF and select m = 3.