Incentivized Communication for Federated Bandits

Authors: Zhepei Wei, Chuanhao Li, Haifeng Xu, Hongning Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive empirical experiments on both synthetic and real-world datasets further validate the effectiveness of the proposed method across various environments.
Researcher Affiliation Academia University of Virginia1 University of Chicago2
Pseudocode Yes Algorithm 1 INC-FEDUCB Algorithm Algorithm 2 Payment-free Incentive Mechanism Algorithm 3 Payment-efficient Incentive Mechanism Algorithm 4 Heuristic Search
Open Source Code No The paper does not provide an explicit statement about open-sourcing the code or a link to a code repository.
Open Datasets Yes We also conduct comprehensive experiments on the real-world recommendation dataset Movie Lens [12].
Dataset Splits No The paper does not explicitly provide training/test/validation dataset splits. It mentions using "synthetic and real-world datasets" but does not detail how these datasets were split for different phases of model development or evaluation.
Hardware Specification No The paper states: "We simulate the incentivized federated bandit problem under various environment settings." This implies a software simulation, and no specific hardware (like GPU/CPU models or cloud instances) is mentioned for running these simulations.
Software Dependencies No The paper does not provide specific version numbers for any software, libraries, or programming languages used in the experiments.
Experiment Setup Yes Specifically, we create an environment of N = 50 clients with cost of data sharing Dp = {Dp 1, , Dp N}, total number of iterations T = 5, 000, feature dimension d = 25, and time-varing arm pool size K = 25. By default, we set Dp i = Dp R, i [N]. Specifically, we explored different hyper-parameter settings for INC-FEDUCB with determinant ratio threshold β [0.3, 0.7, 1], and various environmental configurations with data sharing cost Dp [1, 10, 100].