2022/03/15 by Yinsong Wang, Yu Ding, Wang, Yinsong +3 · 1 citation
Computer Science · Medicine · #Advanced Neural Network Applications #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Medical Imaging Techniques and Applications #Signal Processing (eess.SP) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2203.08317
openalex publication_date 2022/03/15 · openalex created_date 2022/05/05 · openalex updated_date 2026/07/28
Real-time density estimation is ubiquitous in many applications, including\ncomputer vision and signal processing. Kernel density estimation is arguably\none of the most commonly used density estimation techniques, and the use of\n"sliding window" mechanism adapts kernel density estimators to dynamic\nprocesses. In this paper, we derive the asymptotic mean integrated squared\nerror (AMISE) upper bound for the "sliding window" kernel density estimator.\nThis upper bound provides a principled guide to devise a novel estimator, which\nwe name the temporal adaptive kernel density estimator (TAKDE). Compared to\nheuristic approaches for "sliding window" kernel density estimator, TAKDE is\ntheoretically optimal in terms of the worst-case AMISE. We provide numerical\nexperiments using synthetic and real-world datasets, showing that TAKDE\noutperforms other state-of-the-art dynamic density estimators (including those\noutside of kernel family). In particular, TAKDE achieves a superior test\nlog-likelihood with a smaller runtime.\n