Collaborative Learning

Resilient Collaborative Machine Learning via Network-aware Knowledge Consensus

Novelty

• Knowledge similarity measure
• New convergence analysis
• Plug-and-play collaborative learning framework
• Knowledge support vectors.
  

Advantages

• Consensus-based learning makes the proposed approach more natural for NTNs, since network communication is adaptive and significantly reduced.
• The plug-and-play framework increase scalability and resiliency to extreme events, including the complete loss of agents.

Testbed

Relevant paper

• Large-Scale Resilient Collaborative Machine Learning
• Concurrent Learning Adaptive Model Predictive Control with Pseudospectral Implementation
• Robust_Consensus Control for Leader-follower Networked System