Federated Generalized Bayesian Learning via Distributed Stein Variational Gradient Descent
This addresses federated learning challenges by offering a flexible Bayesian method for distributed settings, though it appears incremental as an extension of existing variational gradient descent techniques.
The paper tackles the problem of federated learning by proposing Distributed Stein Variational Gradient Descent (DSVGD), a non-parametric generalized Bayesian inference framework that uses particles to represent the global posterior, enabling a trade-off between communication load and rounds. It shows favorable accuracy and scalability compared to benchmarks while providing well-calibrated predictions.
This paper introduces Distributed Stein Variational Gradient Descent (DSVGD), a non-parametric generalized Bayesian inference framework for federated learning. DSVGD maintains a number of non-random and interacting particles at a central server to represent the current iterate of the model global posterior. The particles are iteratively downloaded and updated by one of the agents with the end goal of minimizing the global free energy. By varying the number of particles, DSVGD enables a flexible trade-off between per-iteration communication load and number of communication rounds. DSVGD is shown to compare favorably to benchmark frequentist and Bayesian federated learning strategies, also scheduling a single device per iteration, in terms of accuracy and scalability with respect to the number of agents, while also providing well-calibrated, and hence trustworthy, predictions.