https://github.com/rickyxume/fed_graph

CIKM-CUP 2022（rank 6/1746）
https://github.com/rickyxume/fed_graph

federated-learning graph-neural-networks

Last synced: 5 days ago
JSON representation

CIKM-CUP 2022（rank 6/1746）

• Host: GitHub
• URL: https://github.com/rickyxume/fed_graph
• Owner: rickyxume
• License: apache-2.0
• Created: 2022-09-11T22:21:13.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2022-09-12T05:03:44.000Z (over 2 years ago)
• Last Synced: 2023-09-10T15:58:12.419Z (over 1 year ago)
• Topics: federated-learning, graph-neural-networks
• Language: Python
• Homepage:
• Size: 407 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
Team name:

联邦对不队

1. Preparing the federatedscope environment

Follow the official tutorial to build the federatedscope environment.

Our team's code execution environment is Centos, GPU is V100 with 32G RAM.

2. To run the code:

> python federatedscope/main.py --cfg federatedscope/gfl/baseline/final_a.yaml --client_cfg federatedscope/gfl/baseline/final.yaml

3. Brief introduction of the developed algorithm:

All clients are trained based on the GIN model by isolated training (no federated learning methods are used).

The best model for each clietnt will be saved according to val_improtve_ratio.

References:

> GIN refer from Graph Isomorphism Network model from the "How Powerful are Graph Neural Networks?" paper, in ICLR'19