Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Multiplayer Federated Learning: Reaching Equilibrium with Less Communication
Authors: TaeHo Yoon, Sayantan Choudhury, Nicolas Loizou
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we verify our theory through numerical experiments. ... In this section, we conduct experiments to assess the empirical performance of PEARL-SGD and verify our theory. We focus on two setups: a multiplayer game with quadratic objectives, and a distributed mobile robot control problem. |
| Researcher Affiliation | Academia | Tae Ho Yoon Sayantan Choudhury Nicolas Loizou Department of Applied Mathematics & Statistics Mathematical Institute for Data Science Johns Hopkins University EMAIL |
| Pseudocode | Yes | Algorithm 1 Per-Player Local SGD (PEARL-SGD) |
| Open Source Code | Yes | We will include the source codes with instructions in the supplemental material. |
| Open Datasets | No | The paper describes generating synthetic data for a "Quadratic n-player game" and using parameter values from a cited work for "Distributed mobile robot control". Neither of these constitutes a publicly available dataset in the conventional sense of a pre-existing collection of data points with explicit access information (link, DOI, etc.). |
| Dataset Splits | No | The paper does not use pre-existing datasets that would typically require explicit training/test/validation splits. For the 'Quadratic n-player game', data is generated, and for the 'Distributed mobile robot control', specific problem parameters are used, rather than a dataset that needs splitting. |
| Hardware Specification | Yes | Experiments were conducted using a personal Mac Book with an Apple M3 chip and 16GB RAM. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers, such as Python versions, libraries, or frameworks used for implementation. |
| Experiment Setup | Yes | We run PEARL-SGD with the theoretical step-size ฮณ = 1/(โฯ+2(ฯ 1)Lmax ฮบ) from Theorems 3.3 and 3.4 with ฯ {1, 2, 4, 5, 8, 20}. ... We set n = 5, d = 10 and M = 100. The matrices Ai,m are generated randomly with their eigenvalues in the range [ยตA, LA] (0 < ยตA < LA). ... We add Gaussian noise with ฯ2 = 100 to the gradients to simulate stochasticity. |