decentralized-gnn

A fully peer-to-peer framework to define and test graph neural network fragments for decentralized devices to classify themselves.

Library

License
Apache License 2.0 (Apache-2.0)

Main Characteristic
  • Fragmented gossip learning or pre-training of base classifiers
  • Decentralized graph diffusion across model fragments
  • Decentralized variation of decoupled graph neural networks
  • Asynchronous and irregular communication
  • Theoretical guarantees
  • Simulations for experimental validation of new algorithms
Research areas
Collaborative AI
Technical Categories
AI services Machine learning Multi-agent systems
Business Categories
Cloud, Edge and Infrastructure
Last updated
16.09.2022 - 09:52

Related Projects
AI4Media