SFedPO: Streaming Federated Learning with a Prediction Oracle under Temporal Shifts

Federated Learning (FL) enables decentralized clients to collaboratively train a global model without sharing raw data. However, most existing FL frameworks assume that clients train on static local datasets collected in advance or that the data follows a fixed underlying distribution, which limits their applicability in dynamic environments where data evolves over time. A parallel line of research, online FL, removes all assumptions and adopts an adversarial perspective, but this approach is often overly pessimistic and neglects the structured, partially predictable nature of real-world data dynamics. To bridge this gap, we propose SFedPO, a streaming federated learning framework that incorporates a prediction oracle to capture the temporal evolution of client-side data distributions. We theoretically analyze the convergence bounds of SFedPO and develop two practical sampling strategies: a Distribution-guided Data Sampling (DDS) strategy that dynamically selects training data under limited storage by balancing historical reuse and distribution adaptation, and a Shift-aware Aggregation Weights (SAW) mechanism that modulates global aggregation based on client-specific sampling behaviors. We further establish robustness guarantees under prediction errors. Extensive experiments demonstrate that SFedPO effectively adapts to streaming scenarios with distribution shifts and significantly outperforms existing methods.

Modern devices like smartphones, vehicles, and sensors continuously generate new data, but most federated learning systems assume that this data stays unchanged over time. In reality, user behavior, environments, and data patterns often evolve, which can quickly make AI models outdated or unreliable. Existing approaches either ignore these changes or assume the worst-case scenario, making them less effective for real-world applications. To address this problem, we propose a new federated learning framework called SFedPO that helps AI models adapt to continuously changing data streams. Our method uses predictions about how each client’s data may evolve over time, allowing the system to better decide which data should be stored and trained on, even when device storage is limited. We also design a mechanism that adjusts how much influence each client has during global model updates based on how their data changes. We provide theoretical guarantees showing that our framework remains reliable even when predictions are imperfect. Experiments on multiple datasets demonstrate that SFedPO adapts effectively to changing environments and achieves significantly better performance than existing federated learning methods.