Streaming Generated Gaussian Process Experts for Online Learning and Control

2025-08-05

AI Summary

Gaussian processes (GPs) suffer from high computational and memory complexity in streaming-data settings, hindering real-time, large-scale deployment. To address this, we propose a bounded-capacity framework that dynamically generates Gaussian process experts via a kernel-induced mechanism, enabling efficient online learning and control with theoretical guarantees. The framework comprises two complementary variants—one prioritizing predictive accuracy and the other emphasizing computational efficiency—both grounded in polynomial-time online algorithms. Extensive experiments on multiple benchmark streaming datasets and real-time control tasks demonstrate that our approach significantly outperforms existing streaming GP methods: it maintains high prediction accuracy while reducing time and space complexity to scalable levels. Crucially, the method is suitable for safety-critical dynamic systems requiring reliable, low-latency inference and adaptation.

Technology Category

• Machine Learning: Online Learning & Bandits
• Reasoning under Uncertainty: Stochastic Optimization
• Search and Optimization: Learning to Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Gaussian Processes (GPs), as a nonparametric learning method, offer flexible modeling capabilities and calibrated uncertainty quantification for function approximations. Additionally, GPs support online learning by efficiently incorporating new data with polynomial-time computation, making them well-suited for safety-critical dynamical systems that require rapid adaptation. However, the inference and online updates of exact GPs, when processing streaming data, incur cubic computation time and quadratic storage memory complexity, limiting their scalability to large datasets in real-time settings. In this paper, we propose a underline{s}treaming underline{k}ernel-induced progressivelunderline{y} generated expert framework of underline{G}aussian underline{p}rocesses (SkyGP) that addresses both computational and memory constraints by maintaining a bounded set of experts, while inheriting the learning performance guarantees from exact Gaussian processes. Furthermore, two SkyGP variants are introduced, each tailored to a specific objective, either maximizing prediction accuracy (SkyGP-Dense) or improving computational efficiency (SkyGP-Fast). The effectiveness of SkyGP is validated through extensive benchmarks and real-time control experiments demonstrating its superior performance compared to state-of-the-art approaches.

Problem

• Address computational complexity of exact GPs for streaming data
• Reduce memory constraints in online GP learning
• Maintain performance guarantees while enabling real-time adaptation

Innovation

• Streaming kernel-induced expert framework for GPs
• Bounded expert set reduces computation and memory
• Two variants optimize accuracy or efficiency