Ctrl K

Catalyst

Morph: Code Supporting the Publication: Dynamic Topology Optimization for Non-IID Data in Decentralized Learning

1
contributor

Description

Catalyst is a lightweight framework to simulate asynchronous federated learning systems with heterogeneous client speeds.

It supports multiple algorithms and is designed for easy extensibility and reproducibility.

Logo of Catalyst
Keywords
asynchronous
Byzantine Robustness
Client Heterogeneity
federated learning
Programming languages
  • Python 90%
  • Jupyter Notebook 9%
  • Other 1%
License
  • MIT
</>Source code
4TU.
Packages

Reference papers

Contributors

BC
Bart Cox

Member of community

4TU