PAM: Improve the Performance of Randomized Online Algorithms With Machine-Learned Predictions in Online Energy Management
IEEE Internet of Things Journal · 2026-11 · 13 (22) · 52801-52809
DOI: 10.1109/jiot.2026.3728840 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Internet of Things Journal, 2026-11. Verfasst von Minxi Feng und 2 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Natural Science Foundation of Anhui Province.
Abstract (Originalsprache)
Recent online algorithms struggle to model uncertain input and improve prediction accuracy in energy management systems. We introduce the probability assignment module (PAM), a novel approach that leverages relationships among sequential prediction errors. PAM enhances the performance of randomized online algorithms without altering their original prediction models or logic. By focusing on monotone predictions (MPs), where consecutive prediction errors are likely to share the same sign, PAM tunes the maximum likelihood point of the probability distribution function (pdf) in the randomized online algorithms and offers a tradeoff between the original and the tuned response by applying weights. We demonstrate the effectiveness of PAM through theoretical proofs and trace-driven experiments on two classical online problems: ski rental and peak shaving when over half of the sequential predictions are MPs. The theoretical proof shows that PAM can achieve a better bounded competitive ratio and offer a smooth tradeoff between robustness and consistency on MPs. The experimental results show that PAM improves the average competitive ratio of randomized online algorithms by up to 18.4% across the evaluated benchmarks, while also identifying suitable parameter settings for $\theta $ and $\beta $ .
Förderung
- National Natural Science Foundation of China
- Natural Science Foundation of Anhui Province
Zitieren
APA
Feng, M., Zhao, X., & Pei, J. (2026). PAM: Improve the Performance of Randomized Online Algorithms With Machine-Learned Predictions in Online Energy Management. IEEE Internet of Things Journal, 13(22), 52801-52809. https://doi.org/10.1109/jiot.2026.3728840
Vancouver
Feng M, Zhao X, Pei J. PAM: Improve the Performance of Randomized Online Algorithms With Machine-Learned Predictions in Online Energy Management. IEEE Internet of Things Journal. 2026;13(22):52801-52809. doi:10.1109/jiot.2026.3728840
BibTeX
@article{Feng2026,
author = {Feng, Minxi and Zhao, Xin and Pei, Jiaming},
title = {PAM: Improve the Performance of Randomized Online Algorithms With Machine-Learned Predictions in Online Energy Management},
journal = {IEEE Internet of Things Journal},
year = {2026},
volume = {13},
number = {22},
pages = {52801--52809},
doi = {10.1109/jiot.2026.3728840}
}
EUVIMED