Federation formation system with bounded agents improves federated learning readiness
FedReady: Bounded Agents for Federation Formation
Artificial IntelligenceMultiagent Systems
Summary
Federated learning lets many different organizations work together on machine learning without sharing all their raw data, but organizing this collaboration takes a lot of preparation. The authors created FedReady, a system that uses limited artificial intelligence agents to help find eligible sites and prepare data and code safely before training starts. Their system checked many simulated medical sites and tasks, showing it can pick the right participants and get everything ready accurately and efficiently. This makes it easier to run federated learning studies without repeating the hard setup work each time.
What this means in practice
- •For healthcare data teams: Automate selection and preparation of hospital sites and patient data before federated learning for medical studies.
- •For enterprise ai platform operators: Integrate bounded agents to manage readiness checks and artifact verification across multiple participant sites in federated AI workflows.
Authors
Ziyue Xu, Holger R. Roth, Chester Chen, Peter Cnudde
Abstract
Federated learning (FL) runtimes standardize aggregation and deployment, but each new study still requires federation formation: selecting task-eligible sites and preparing their local data and task code for federated training or analysis. This preparation recurs when the target changes, while raw records and site-private mappings are excluded from cross-site FL messages. We present FedReady, a system that performs federation formation with bounded LLM agents. Agents propose shared contracts, local adapters, and task code; the runtime owns workflow state and accepts artifacts only after checks tied to their digests. Candidate site discovery prioritizes recall; local eligibility verification checks task support before site admission. Once participating sites and artifacts are fixed, conventional federated training or analysis proceeds without readiness agents. Across 39 simulated retinal sites and four tasks, discovery reaches recall 1.00 at precision 0.65; local eligibility verification removes all 17 false positives without losing an eligible site-task pair. The participating sites supply 45,575 task-record instances and complete four 100-round jobs, including one with a runtime-generated contract. Across ten patient cohorts, preparation supports seven patient populations and holds endpoints at three sources as unsupported. Comparison with manually prepared reference populations and analysis (the manual reference) gives a maximum survival-curve difference of 0.001133. We also analyze 19 development failures, targeted classification ablations, and standalone and heterogeneous agent configurations. These results establish the feasibility of task-specific data preparation before ordinary FL training or analysis.