📐 Learning Theory

🤖 AAAI2026 · 3 paper notes

📌 Same area in other venues: 🔬 ICLR2026 (293) · 🧪 ICML2026 (45) · 🧠 NeurIPS2025 (25) · 🧪 ICML2025 (16)

Streaming Generated Gaussian Process Experts for Online Learning and Control: Extended Version

This paper proposes SkyGP (Streaming Kernel-induced Progressively Generated Expert GP), which handles streaming data via kernel-distance-driven progressive expert generation and time-aware configurable aggregation, inheriting the learning guarantees of exact GP while maintaining bounded computational complexity. SkyGP comprehensively outperforms state-of-the-art methods on both benchmark regression tasks and real-time control experiments.

Generalizing Analogical Inference from Boolean to Continuous Domains

This paper revisits the theoretical foundations of analogical inference: it first constructs a counterexample demonstrating the failure of classical generalization bounds in the Boolean domain, then proposes a unified analogical inference framework based on parameterized generalized means, extending discrete classification to continuous regression domains.

A Switching Framework for Online Interval Scheduling with Predictions

For the irrevocable online interval scheduling problem, this paper proposes the SemiTrust-and-Switch framework and the SmoothMerge randomized algorithm. By switching between or blending a prediction-trusting strategy and a classical greedy algorithm, the approach achieves near-optimal performance when predictions are accurate (consistency) and degrades gracefully when predictions are erroneous (robustness and smoothness). Tightness of the framework on specific instances is also established.