DA-Distill: Dual-Alignment Distillation for Multimodal Knowledge Transfer on Edge Devices
IEEE Transactions on Consumer Electronics · 2025-08 · 71 (3) · 8055-8066
DOI: 10.1109/tce.2025.3598319 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Consumer Electronics, 2025-08. Verfasst von Caiyu Fan und 4 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 47 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities, Young Scientist Program of The National New Energy Vehicle Technology Innovation Center.
Abstract (Originalsprache)
The growing demand for multimodal functionalities in consumer electronics has driven the widespread adoption of multimodal pre-trained models, such as CLIP, enhancing human-computer interactions on edge devices through adaptive and context-aware functionalities. However, deep neural architectures impose significant computational complexity, storage demands, and data privacy challenges that limit traditional fine-tuning methods at the edge. To overcome these constraints, we propose an efficient knowledge transfer framework comprising two main modules: a data filtering module employing a combined confidence-driven and random sampling strategy to select representative samples while ensuring privacy compliance, and a knowledge distillation module that explicitly transfers multimodal representations via contrastive learning. Specifically, we construct bidirectional cross-modal contrastive distributions and minimize their Kullback-Leibler divergence to achieve representation alignment between the old and new models. Experimental results indicate that our method significantly outperforms conventional fine-tuning, achieving more than 1.5% accuracy improvement using only 5% of the original training data. Additionally, we identify optimal data selection strategies tailored to different data availability scenarios, demonstrating the practical effectiveness and flexibility of our approach for resource-constrained edge computing applications.
Förderung
- National Natural Science Foundation of China
- Fundamental Research Funds for the Central Universities
- Young Scientist Program of The National New Energy Vehicle Technology Innovation Center
Zitieren
APA
Fan, C., Song, K., Zou, B., Hu, X., & Ma, H. (2025). DA-Distill: Dual-Alignment Distillation for Multimodal Knowledge Transfer on Edge Devices. IEEE Transactions on Consumer Electronics, 71(3), 8055-8066. https://doi.org/10.1109/tce.2025.3598319
Vancouver
Fan C, Song K, Zou B, Hu X, Ma H. DA-Distill: Dual-Alignment Distillation for Multimodal Knowledge Transfer on Edge Devices. IEEE Transactions on Consumer Electronics. 2025;71(3):8055-8066. doi:10.1109/tce.2025.3598319
BibTeX
@article{Fan2025,
author = {Fan, Caiyu and Song, Kun and Zou, Bochao and Hu, Xiaocheng and Ma, Huimin},
title = {DA-Distill: Dual-Alignment Distillation for Multimodal Knowledge Transfer on Edge Devices},
journal = {IEEE Transactions on Consumer Electronics},
year = {2025},
volume = {71},
number = {3},
pages = {8055--8066},
doi = {10.1109/tce.2025.3598319}
}
EUVIMED