Scalable Decentralized Algorithms for Online Personalized Mean Estimation

Authors: Franco Galante, Giovanni Neglia, Emilio Leonardi

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct preliminary experiments demonstrating how our algorithms can be adapted to federatedly learn more general machine learning models. Section 6: Numerical Experiments. Figure 1 showcases the performance of B-Col ME and C-Col ME using two key metrics: the fraction of agents with incorrect estimates (ˆµt a more than ε away from the true mean µa), and the fraction of wrong links still in use.
Researcher Affiliation Academia 1Politecnico di Torino 2INRIA EMAIL, EMAIL, EMAIL
Pseudocode Yes Algorithm 1: C-Col ME over a Time Horizon H; Algorithm 2: B-Col ME over a Time Horizon H
Open Source Code Yes Code https://github.com/Franco-Galante/scalable-decentralized-algorithms-AAAI25
Open Datasets Yes We adapt our approach for personalized federated learning, applying it to the task of handwritten digit recognition using the MNIST dataset. (Deng 2012)
Dataset Splits No The paper mentions generating data from Gaussian distributions and adapting the approach for handwritten digit recognition using the MNIST dataset (Deng 2012). It describes how clients receive new data samples from one of two distributions, and how the MNIST dataset is used by swapping/maintaining labels. However, it does not explicitly provide specific training/test/validation dataset splits (e.g., percentages or sample counts) for reproducibility of the experiments.
Hardware Specification No Computational resources provided by hpc@polito, which is a project of Academic Computing within the Department of Control and Computer Engineering at the Politecnico di Torino (http://hpc.polito.it). This text mentions a computing resource but does not specify any particular hardware like GPU models, CPU models, or detailed computer specifications.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python 3.8, PyTorch 1.9, etc.).
Experiment Setup Yes Unless otherwise stated, in the experiments |A| = N = 10000, r = 10, d = 4, ε = 0.1, δ = 0.1, and βγ(n) as in (3).