Communication Efficient Federated Learning for Generalized Linear Bandits

Authors: Chuanhao Li, Hongning Wang

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We rigorously proved, though the setting is more general and challenging, our algorithm can attain sub-linear rate in both regret and communication cost, which is also validated by our extensive empirical evaluations. ... Extensive empirical evaluations on both synthetic and real-world datasets are performed to validate the effectiveness of our algorithm.
Researcher Affiliation Academia 1Department of Computer Science, University of Virginia
Pseudocode Yes Algorithm 1 Fed GLB-UCB; Algorithm 2 ONS-Update; Algorithm 3 AGD-Update
Open Source Code Yes Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [Yes]
Open Datasets Yes Real-world Dataset ... Cover Type, Magic Telescope and Mushroom from the UCI Machine Learning Repository [5]. [5] Dheeru Dua and Casey Graff. UCI machine learning repository, 2017.
Dataset Splits No The paper describes parameters for the synthetic and real-world datasets (e.g., T=2000, N=200), and how data is generated/collected online in a bandit setting, but it does not specify explicit train/validation/test dataset splits with percentages or counts, which are typical for supervised learning reproducibility.
Hardware Specification No Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [No]
Software Dependencies No The paper does not provide specific names and version numbers of software dependencies (e.g., programming languages, libraries, frameworks, or solvers) used for the experiments.
Experiment Setup Yes We simulated the federated GLB setting defined in Section 3.3, with T = 2000, N = 200, d = 10, S = 1, At (K = 25) uniformly sampled from a ℓ2 unit sphere, and reward yt,,i Bernoulli(µ(x t,,iθ )), with µ(z) = (1 + exp( z)) 1. ... we pre-processed these datasets following the steps in prior works [8], with T = 2000 and N = 20. ... we run Fed GLB-UCB with different threshold value D (logarithmically spaced between 10 1 and 103) and its variants with different number of global updates B.