LINDT: Tackling Negative Federated Learning with Local Adaptation
This work addresses a critical failure mode in federated learning for practitioners deploying FL systems, offering a method for detection and recovery.
This paper defines and analyzes "negative federated learning" (NFL), a state where federated learning fails to proceed properly. The authors propose LINDT, a framework that detects NFL from the server and recovers by adapting the federated model on each client's local data using a Layer-wise Intertwined Dual-model, significantly improving FL performance on local data.
Federated Learning (FL) is a promising distributed learning paradigm, which allows a number of data owners (also called clients) to collaboratively learn a shared model without disclosing each client's data. However, FL may fail to proceed properly, amid a state that we call negative federated learning (NFL). This paper addresses the problem of negative federated learning. We formulate a rigorous definition of NFL and analyze its essential cause. We propose a novel framework called LINDT for tackling NFL in run-time. The framework can potentially work with any neural-network-based FL systems for NFL detection and recovery. Specifically, we introduce a metric for detecting NFL from the server. On occasion of NFL recovery, the framework makes adaptation to the federated model on each client's local data by learning a Layer-wise Intertwined Dual-model. Experiment results show that the proposed approach can significantly improve the performance of FL on local data in various scenarios of NFL.