EUVIMEDEUVIMED entdecken
Crossref

Decouple Graph Neural Networks: Train Multiple Simple GNNs Simultaneously Instead of One

Hongyuan Zhang, Yanan Zhu, Xuelong Li

IEEE Transactions on Pattern Analysis and Machine Intelligence · 2024-11 · 46 (11) · 7451-7462

DOI: 10.1109/tpami.2024.3392782 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024-11. Verfasst von Hongyuan Zhang und 2 weiteren Autor:innen. Die Arbeit wurde bisher 42-mal zitiert (Crossref) und nennt 58 Literaturangaben. Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Graph neural networks (GNN) suffer from severe inefficiency due to the exponential growth of node dependency with the increase of layers. It extremely limits the application of stochastic optimization algorithms so that the training of GNN is usually time-consuming. To address this problem, we propose to decouple a multi-layer GNN as multiple simple modules for more efficient training, which is comprised of classical forward training (FT) and designed backward training (BT). Under the proposed framework, each module can be trained efficiently in FT by stochastic algorithms without distortion of graph information owing to its simplicity. To avoid the only unidirectional information delivery of FT and sufficiently train shallow modules with the deeper ones, we develop a backward training mechanism that makes the former modules perceive the latter modules, inspired by the classical backward propagation algorithm. The backward training introduces the reversed information delivery into the decoupled modules as well as the forward information delivery. To investigate how the decoupling and greedy training affect the representational capacity, we theoretically prove that the error produced by linear modules will not accumulate on unsupervised tasks in most cases. The theoretical and experimental results show that the proposed framework is highly efficient with reasonable performance, which may deserve more investigation.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Zhang, H., Zhu, Y., & Li, X. (2024). Decouple Graph Neural Networks: Train Multiple Simple GNNs Simultaneously Instead of One. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(11), 7451-7462. https://doi.org/10.1109/tpami.2024.3392782

Vancouver

Zhang H, Zhu Y, Li X. Decouple Graph Neural Networks: Train Multiple Simple GNNs Simultaneously Instead of One. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2024;46(11):7451-7462. doi:10.1109/tpami.2024.3392782

BibTeX
@article{Zhang2024,
  author = {Zhang, Hongyuan and Zhu, Yanan and Li, Xuelong},
  title = {Decouple Graph Neural Networks: Train Multiple Simple GNNs Simultaneously Instead of One},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year = {2024},
  volume = {46},
  number = {11},
  pages = {7451--7462},
  doi = {10.1109/tpami.2024.3392782}
}
Kostenlos starten →