Sparse Uncertainty-Informed Sampling from Federated Streaming Data
This is an incremental improvement for federated learning systems dealing with streaming data under resource constraints.
The paper tackles the problem of sampling non-I.I.D. data streams in federated learning with limited resources and sparse labeled data, resulting in enhanced training batch diversity and improved numerical robustness compared to existing strategies.
We present a numerically robust, computationally efficient approach for non-I.I.D. data stream sampling in federated client systems, where resources are limited and labeled data for local model adaptation is sparse and expensive. The proposed method identifies relevant stream observations to optimize the underlying client model, given a local labeling budget, and performs instantaneous labeling decisions without relying on any memory buffering strategies. Our experiments show enhanced training batch diversity and an improved numerical robustness of the proposal compared to existing strategies over large-scale data streams, making our approach an effective and convenient solution in FL environments.