EUVIMEDEUVIMED entdecken
Crossref

Dual-balancing for multi-task learning

Baijiong Lin, Weisen Jiang, Feiyang Ye, Yu Zhang, Pengguang Chen, Ying-Cong Chen, Shu Liu, Ivor W Tsang, James T Kwok

Neural Networks · 2026-03 · 195 · 108317

DOI: 10.1016/j.neunet.2025.108317 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Neural Networks, 2026-03. Verfasst von Baijiong Lin und 8 weiteren Autor:innen. Beteiligte Einrichtung: The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, 510000, China; HKUST(GZ) - SmartMore Joint Lab, Guangzhou, 510000, China. …. Hauptthemen laut MEDLINE: Learning, Machine Learning, Neural Networks, Computer. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 61 Literaturangaben. Gefördert von: National Natural Science Foundation of China.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Multi-task learning aims to learn multiple related tasks simultaneously and has achieved great success in various fields. However, the disparity in loss and gradient scales among tasks often leads to performance compromises, and the balancing of tasks remains a significant challenge. In this paper, we propose Dual-Balancing Multi-Task Learning (DB-MTL) to achieve task balancing from both the loss and gradient perspectives. Specifically, DB-MTL achieves loss-scale balancing by performing logarithm transformation on each task loss, and rescales gradient magnitudes by normalizing all task gradients to comparable magnitudes using the maximum gradient norm. Extensive experiments on a number of benchmark datasets demonstrate that DB-MTL consistently performs better than the current state-of-the-art.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

LearningMachine LearningNeural Networks, ComputerHumansAlgorithms

Stichwörter der Autor:innen

Gradient balancingLoss balancingMulti-task learning

Autor:innen und Institutionen

  • The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, 510000, China; HKUST(GZ) - SmartMore Joint Lab, Guangzhou, 510000, China. Electronic address: bj.lin.email@gmail.com.
  • The Chinese University of Hong Kong, Hong Kong, 999077, China.
  • Southern University of Science and Technology, Shenzhen, 518055, China.
  • SmartMore, Shenzhen, 518000, China.
  • The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, 510000, China; HKUST(GZ) - SmartMore Joint Lab, Guangzhou, 510000, China; The Hong Kong University of Science and Technology, Hong Kong, 999077,…
  • SmartMore, Shenzhen, 518000, China. Electronic address: liushuhust@gmail.com.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Lin, B., Jiang, W., Ye, F., Zhang, Y., Chen, P., Chen, Y. C., Liu, S., Tsang, I. W., & Kwok, J. T. (2026). Dual-balancing for multi-task learning. Neural Networks, 195, 108317. https://doi.org/10.1016/j.neunet.2025.108317

Vancouver

Lin B, Jiang W, Ye F, Zhang Y, Chen P, Chen YC, et al. Dual-balancing for multi-task learning. Neural Networks. 2026;195:108317. doi:10.1016/j.neunet.2025.108317

BibTeX
@article{Lin2026,
  author = {Lin, Baijiong and Jiang, Weisen and Ye, Feiyang and Zhang, Yu and Chen, Pengguang and Chen, Ying-Cong and Liu, Shu and Tsang, Ivor W and Kwok, James T},
  title = {Dual-balancing for multi-task learning},
  journal = {Neural Networks},
  year = {2026},
  volume = {195},
  pages = {108317},
  doi = {10.1016/j.neunet.2025.108317}
}
Kostenlos starten →