Personalized and Resilient Distributed Learning Through Opinion Dynamics
Luca Ballotta, Nicola Bastianello, Riccardo M. G. Ferrari, Karl H. Johansson
TL;DR
The paper addresses personalization and resilience in distributed learning by marrying distributed gradient descent with Friedkin-Johnsen opinion dynamics, controlled by a stubbornness parameter λ. It proves linear convergence to a fixed point that blends global optimum and local optima, and extends to scenarios with bounded adversarial updates, offering robustness through partial collaboration. Extensive experiments on synthetic classification tasks and MNIST demonstrate improved local personalization without sacrificing, and often improving, global accuracy, while mitigating malicious influence. The approach is lightweight, does not rely on dense connectivity or trusted nodes, and provides practical tunability for different network heterogeneity and security conditions.
Abstract
In this paper, we address two practical challenges of distributed learning in multi-agent network systems, namely personalization and resilience. Personalization is the need of heterogeneous agents to learn local models tailored to their own data and tasks, while still generalizing well; on the other hand, the learning process must be resilient to cyberattacks or anomalous training data to avoid disruption. Motivated by a conceptual affinity between these two requirements, we devise a distributed learning algorithm that combines distributed gradient descent and the Friedkin-Johnsen model of opinion dynamics to fulfill both of them. We quantify its convergence speed and the neighborhood that contains the final learned models, which can be easily controlled by tuning the algorithm parameters to enforce a more personalized/resilient behavior. We numerically showcase the effectiveness of our algorithm on synthetic and real-world distributed learning tasks, where it achieves high global accuracy both for personalized models and with malicious agents compared to standard strategies.