A multi-lag generative model and inference for long-term dependencies in temporal point processes
Electronic Journal of Statistics · 01.01.2026 · 20 (2) · 5669-5704
DOI: 10.1214/26-ejs2611 ↗Kurz gefasst
Fachartikel in der Zeitschrift Electronic Journal of Statistics, erschienen am 01.01.2026. Verfasst von Muhong Gao und 1 weiteren Autor:innen. Die Arbeit nennt 60 Literaturangaben.
Abstract (Originalsprache)
Network-structured temporal point processes (TPPs) are widely used in scientific applications, yet statistical learning remains challenging due to the complex stochastic mechanisms governed by conditional intensity functions (CIFs). Existing methods for uncovering interactions among V nodes primarily rely on short-lag predictive models of CIFs and are often limited to single-lag dependencies on past events. To address this limitation, we propose a multi-lag generative model and statistical inference approach for unsupervised learning of CIFs from temporal event streams. Our continuous-time model explicitly encodes K-lag dependencies over a finite history horizon spanning multiple lag windows, yielding an interpretable and flexible representation of event dynamics across multiple time scales. To estimate the resulting high-dimensional model with V2K interaction parameters and V baseline parameters, we develop a penalized learning procedure that accommodates diverse link and loss functions and provides theoretical guarantees for network structure recovery. Experiments on synthetic data and real-world datasets, including neuronal spike trains and crime records, demonstrate that the proposed method achieves better performance than the competing methods in both structure learning and forecasting accuracy.
Zitieren
APA
Gao, M., & Zhang, C. (2026). A multi-lag generative model and inference for long-term dependencies in temporal point processes. Electronic Journal of Statistics, 20(2), 5669-5704. https://doi.org/10.1214/26-ejs2611
Vancouver
Gao M, Zhang C. A multi-lag generative model and inference for long-term dependencies in temporal point processes. Electronic Journal of Statistics. 2026;20(2):5669-5704. doi:10.1214/26-ejs2611
BibTeX
@article{Gao2026,
author = {Gao, Muhong and Zhang, Chunming},
title = {A multi-lag generative model and inference for long-term dependencies in temporal point processes},
journal = {Electronic Journal of Statistics},
year = {2026},
volume = {20},
number = {2},
pages = {5669--5704},
doi = {10.1214/26-ejs2611}
}
EUVIMED