Federated Bilevel Performative Prediction

Federated bilevel optimization is widely used for nested learning problems across distributed clients, such as federated hyperparameter tuning and meta-learning under privacy and communication constraints. Most existing formulations assume fixed client data distributions, which can be violated by performativity, where deployed decisions reshape client behavior and data collection, inducing client-specific, decision-dependent distribution shift. We study federated bilevel performative prediction, where both upper-level (UL) and lower-level (LL) objectives are evaluated under client-dependent, decision-dependent distributions. We formalize the federated bilevel performatively stable (FBPS) point under a decoupled-risk perspective and provide sufficient conditions for its existence and uniqueness. We then develop two federated methods to compute the FBPS solution: FBi-RRM, which converges linearly under a contraction condition, and FBi-SGD, a communication-efficient stochastic method based on federated hypergradient estimation with convergence guarantees under diminishing step sizes when sensitivities are sufficiently small. Experiments on strategic regression and meta strategic classification validate the predicted stability thresholds and demonstrate improved meta-generalization over non-performative baselines, and CNN-based classification further demonstrates the practical effectiveness of the proposed methods in nonconvex neural network settings.

Machine learning models are often trained as if the world stays the same after deployment. In many real systems, however, this is not true: once a model is used, people or devices may react to its decisions, and the data collected later can change. This feedback effect is especially challenging in federated learning, where many clients, such as hospitals, banks, or mobile devices, train a shared model without pooling their raw data. It becomes even harder when the learning problem has two connected levels, such as tuning a shared model while each client also adapts its own local model. This paper studies this setting and asks how federated bilevel learning can remain stable when each client’s data distribution changes in response to deployed decisions. We introduce a stability notion called the federated bilevel performatively stable point, which describes a model that remains appropriate after deployment and retraining. We then design two algorithms, FBi-RRM and FBi-SGD, to find this stable solution with theoretical convergence guarantees. Experiments on strategic learning tasks and federated MNIST show that accounting for these feedback effects can improve stability and prediction performance compared with methods that ignore them. This work helps make federated learning more reliable in real-world systems where model decisions and future data influence each other.