Gossip Learning

Gossip learning is a fully decentralized machine learning approach where nodes directly exchange and aggregate model updates without a central server, aiming to improve efficiency and privacy compared to centralized methods like federated learning. Current research focuses on addressing challenges like the "vanishing variance" problem during model averaging and optimizing energy efficiency in resource-constrained environments, often employing adaptive optimization and novel averaging algorithms. This approach holds significant promise for applications in distributed systems, particularly in areas like Internet of Things (IoT) devices and collaborative medical imaging analysis, where data privacy and resource limitations are paramount.

Papers

April 18, 2024

Context-Aware Orchestration of Energy-Efficient Gossip Learning Schemes
Mina Aghaei Dinani, Adrian Holzer, Hung Nguyen, Marco Ajmone Marsan, Gianluca Rizzo
Adaptive Optimization Convergence Orchestration Gossip Learning IoT Node Network Orchestrator

April 6, 2024

Vanishing Variance Problem in Fully Decentralized Neural-Network Systems
Yongding Tian, Zaid Al-Ars, Maksim Kitsak, Peter Hofstee
Model Convergence Model Averaging Gossip Learning Neural Network System

January 27, 2024

Decentralized Gossip Mutual Learning (GML) for brain tumor segmentation on multi-parametric MRI
Jingyun Chen, Yading Yuan
Brain Tumor Segmentation Model Training Mutual Learning Gossip Algorithm Multi Parametric MRI Gossip Learning

January 17, 2024

Technical Report: On the Convergence of Gossip Learning in the Presence of Node Inaccessibility
Tian Liu, Yue Cui, Xueyang Hu, Yecheng Xu, Bo Liu
Early Stage Convergence Speech Presence Technical Report Wireless Network Dynamic Network UAV Network Malicious Node Gossip Learning

January 11, 2024

Decentralized Gossip Mutual Learning (GML) for automatic head and neck tumor segmentation
Jingyun Chen, Yading Yuan
Tumor Segmentation Human Head Global Model Mutual Learning Neck Tumor Segmentation Gossip Learning

June 15, 2023

On the resilience of Collaborative Learning-based Recommender Systems Against Community Detection Attack
Yacine Belal, Sonia Ben Mokhtar, Mohamed Maouche, Anthony Simonet-Boulogne
Differential Privacy Community Detection Tiny Refinement Elicit Resilience Gossip Learning Learning Based Recommender System

February 16, 2023

On the Limit Performance of Floating Gossip
Gianluca Rizzo, Noelia Perez Palma, Marco Ajmone Marsan, Vincenzo Mancuso
Machine Learning Model Continuous Learning Gossip Learning Performance Limit

December 14, 2022

FLAGS Framework for Comparative Analysis of Federated Learning Algorithms
Ahnaf Hannan Lodhi, Barış Akgün, Öznur Özkasap
Comparative Study Federated Averaging Decentralized Federated Learning Federated Learning Algorithm Gossip Learning

August 9, 2022

PEPPER: Empowering User-Centric Recommender Systems over Gossip Learning
Yacine Belal, Aurélien Bellet, Sonia Ben Mokhtar, Vlad Nitu
Recommender System Gossip Algorithm Gossip Learning Decentralized Recommender System

November 29, 2021

p2pGNN: A Decentralized Graph Neural Network for Node Classification in Peer-to-Peer Networks
Emmanouil Krasanakis, Symeon Papadopoulos, Ioannis Kompatsiaris
Graph Neural Network Node Classification Graph Diffusion Peer to Peer Gossip Learning Decentralized Graph Decentralized Graph Neural Network

Gossip Learning

Papers

Context-Aware Orchestration of Energy-Efficient Gossip Learning Schemes

Vanishing Variance Problem in Fully Decentralized Neural-Network Systems

Decentralized Gossip Mutual Learning (GML) for brain tumor segmentation on multi-parametric MRI

Technical Report: On the Convergence of Gossip Learning in the Presence of Node Inaccessibility

Decentralized Gossip Mutual Learning (GML) for automatic head and neck tumor segmentation

On the resilience of Collaborative Learning-based Recommender Systems Against Community Detection Attack

On the Limit Performance of Floating Gossip

FLAGS Framework for Comparative Analysis of Federated Learning Algorithms

PEPPER: Empowering User-Centric Recommender Systems over Gossip Learning

p2pGNN: A Decentralized Graph Neural Network for Node Classification in Peer-to-Peer Networks