Fuzzy Message Passing in Graph Neural Networks: A First Approach to Uncertainty in Node Embeddings

Minh Tuan Duong

doi:10.4108/eetcasa.8947

Authors

Minh Tuan Duong Trường ĐH Nguyễn Tất Thành

DOI:

https://doi.org/10.4108/eetcasa.8947

Keywords:

Graph Neural Networks, Message Passing, Uncertainty, Node Embeddings, Max-Min Aggregation, Fuzzy Logic

Abstract

Graph Neural Networks (GNNs) have emerged as a powerful tool for learning representations in graph structured data. However, traditional message-passing mechanisms often struggle with uncertainty and noise in node features and graph topology. In this paper, we propose Fuzzy Message Passing (FMP), a novel approach that integrates fuzzy max-min aggregation into GNNs to improve robustness against uncertainty. Our method enhances node embeddings by leveraging fuzzy logic principles, ensuring better stability and interpretability in complex graph tasks. Experimental results on benchmark datasets demonstrate that FMP outperforms conventional message-passing schemes, particularly in scenarios with noisy or incomplete data.

References

[1] William Hamilton, Rex Ying, and Jure Leskovec. Inductive representation learning on large graphs. In Advances in Neural Information Processing Systems (NeurIPS), 2017.

[2] Nguyen Van Han and Nguyen Cong Hao. Algorithm for reasoning with words based on linguistic fuzzy cognitive map. Concurrency and Computation Practice and Experience, 35(15):e6415, may 2021.

[3] Nguyen Van Han and Phan Cong Vinh. Reasoning with words: A hedge algebra linguistic cognitive map approach. Concurrency and Computation Practice and Experience, 33(2):e5711, Feb 2020.

[4] Nguyen Van Han and Phan Cong Vinh. Toward modeling fuzzy dynamic system based on linguistic values. Mobile Networks and Applications, 26(3):1339–1343, 2021.

[5] Nguyen Cat Ho, Tran Thai Son, Tran Dinh Khang, and Le Xuan Viet. Fuzziness measure, quantified semantic mapping and interpolative method of approximate reasoning in medical expert systems. Journal of Computer Science and Cybernetics, 18(3):237–252, 2002.

[6] Nguyen Cat Ho and W.Wechler. Hedge algebras: An algebraic approach to structure of sets of linguistic truth values. Fuzzy Sets and Systems, 35(3):281–293, 1990.

[7] Thomas N. Kipf and Max Welling. Semi-supervised classification with graph convolutional networks. International Conference on Learning Representations (ICLR), 2017.

[8] L.A.Zadeh. Fuzzy sets. Information and Control, 8:338–353, 1965.

[9] Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. Graph attention networks. In International Conference on Learning Representations (ICLR), 2018.

[10] Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and S. Yu Philip. A comprehensive survey on graph neural networks. IEEE Transactions on Neural Networks and Learning Systems, 2020.

[11] Chao Zhang, Yang Song, Chong Huang, Filip Swiderski, and Paul Kingsbury. Bayesian graph neural networks for semi-supervised classification. In Advances in Neural Information Processing Systems (NeurIPS), 2019.