LINDT: Tackling Negative Federated Learning with Local Adaptation

TL;DR

This paper introduces LINDT, a framework for detecting and recovering from negative federated learning (NFL) in real-time, improving model performance by local adaptation using a novel dual-model approach.

Contribution

The paper formulates a rigorous definition of NFL, proposes a detection metric, and introduces a novel local adaptation framework called LINDT for NFL recovery.

Findings

LINDT effectively detects NFL using a new metric.

LINDT significantly improves FL performance in NFL scenarios.

The dual-model adaptation enhances local data learning during NFL.

Abstract

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.