What is it about?
We study how intelligence spreads among uniform users that randomly interact with each other. This study is the first one that shows the negative impact of dataset fragmentation across several devices on intelligence and learning speed through extensive experimental validation on two very popular datasets for handwritten digit and object-recognition tasks.
Featured Image
Photo by Growtika on Unsplash
Why is it important?
Before this study, no other had experimentally quantified the accuracy reduction and the variation in convergence time due to the separate learning and intelligence exchange among users via a Monte Carlo simulation on random networks.
Perspectives
This article explores the performance of intelligence propagation over networked participants from a systems perspective and provides high-confidence evidence of the effects of network dynamics on learning performance.
Antonio Di Maio
University of Bern
Read the Original
This page is a summary of: The Upsides of Turbulence: Baselining Gossip Learning in Dynamic Settings, October 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3565287.3616530.
You can read the full text:
Contributors
The following have contributed to this page
