Passive Party

Passive party participation in collaborative machine learning and multi-agent systems is a growing research area focused on incentivizing and securing contributions from parties lacking complete data or labels. Current research explores methods like incentive allocation based on game theory (e.g., bankruptcy games and Shapley values) and robust model architectures that mitigate performance degradation and intellectual property leakage when passive parties unexpectedly withdraw. These advancements are crucial for enabling secure and efficient collaboration in federated learning and other distributed applications, particularly in scenarios with sensitive data or unreliable participants.