Newest papers

The Dynamics of Collapse and Recovery in Clustered Federated Learning

Priyanka Deb Sinha, Snigdhansu Chatterjee · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28565 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Em particular, sob inicialização desfavorável, as atribuições dos primeiros clientes concentram grupos heterogêneos em um modelo de cluster, que é então treinado em direção a um compromisso combinado, enquanto outro modelo não recebe atualizações. Uma vez que os modelos atuais determinam o roteamento do cliente e o roteamento determinam as atualizações subsequentes do modelo, formulamos o aprendizado federado do cluster como controle de um sistema dinâmico desconhecido, com modelos de cluster como o estado em evolução, padrões de roteamento do cliente como a entrada de controle, e intervenções de roteamento como controladores autônomos dependentes de estado. Em um modelo linear de dois grupos, identificamos uma região de inicialização em que o roteamento ganancioso se torna auto-enforcado. Em seguida, derivamos a dinâm

Machine translation (Meta M2M100); the English original follows.

Abstract

Clustered federated learning (CFL) can suffer from cluster collapse. In particular, under unfavorable initialization, early client assignments concentrate heterogeneous groups on one cluster model, which is then trained toward a pooled compromise, while another model receives no updates. Since the current models determine client routing and routing determines subsequent model updates, we formulate clustered federated learning as control of an unknown dynamical system, with cluster models as the evolving state, client routing patterns as the control input, and routing interventions as state dependent controllers. In a two group linear model, we identify an initialization region in which greedy routing makes collapse self reinforcing. We then derive finite horizon error dynamics showing how routing interventions can reverse collapse by repeatedly training the starved model, primarily with updates from its compatible group. Motivated by this condition, each client uses blocked EXP3 to choose between greedy routing and interventions that redirect updates toward unused models. Experiments show that with simpler predictors and faster specialist formation, autonomous controller selection recovers usable specialists from unfavorable starts that trap standard CFL baselines. In transformed FEMNIST, the synchronized-entry configuration with persistent commitment recovers usable specialists in settings where the autonomous controller configuration often does not. Code and scripts for reproducing the experiments and figures are available at https://github.com/prsinha1/FedRecover.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Routing (electronic design automation), Initialization, Distributed computing, USable, Static routing

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange