Sparse Causal Discovery with Generative Intervention for Unsupervised Graph Domain Adaptation

Junyu Luo, Yuhao Tang, Yiwei Fu, Xiao Luo, Zhizhuo Kou, Zhiping Xiao, Wei Ju, Wentao Zhang, Ming Zhang
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:41331-41345, 2025.

Abstract

Unsupervised Graph Domain Adaptation (UGDA) leverages labeled source domain graphs to achieve effective performance in unlabeled target domains despite distribution shifts. However, existing methods often yield suboptimal results due to the entanglement of causal-spurious features and the failure of global alignment strategies. We propose SLOGAN (Sparse Causal Discovery with Generative Intervention), a novel approach that achieves stable graph representation transfer through sparse causal modeling and dynamic intervention mechanisms. Specifically, SLOGAN first constructs a sparse causal graph structure, leveraging mutual information bottleneck constraints to disentangle sparse, stable causal features while compressing domain-dependent spurious correlations through variational inference. To address residual spurious correlations, we innovatively design a generative intervention mechanism that breaks local spurious couplings through cross-domain feature recombination while maintaining causal feature semantic consistency via covariance constraints. Furthermore, to mitigate error accumulation in target domain pseudo-labels, we introduce a category-adaptive dynamic calibration strategy, ensuring stable discriminative learning. Extensive experiments on multiple real-world datasets demonstrate that SLOGAN significantly outperforms existing baselines.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-luo25j, title = {Sparse Causal Discovery with Generative Intervention for Unsupervised Graph Domain Adaptation}, author = {Luo, Junyu and Tang, Yuhao and Fu, Yiwei and Luo, Xiao and Kou, Zhizhuo and Xiao, Zhiping and Ju, Wei and Zhang, Wentao and Zhang, Ming}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {41331--41345}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/luo25j/luo25j.pdf}, url = {https://proceedings.mlr.press/v267/luo25j.html}, abstract = {Unsupervised Graph Domain Adaptation (UGDA) leverages labeled source domain graphs to achieve effective performance in unlabeled target domains despite distribution shifts. However, existing methods often yield suboptimal results due to the entanglement of causal-spurious features and the failure of global alignment strategies. We propose SLOGAN (Sparse Causal Discovery with Generative Intervention), a novel approach that achieves stable graph representation transfer through sparse causal modeling and dynamic intervention mechanisms. Specifically, SLOGAN first constructs a sparse causal graph structure, leveraging mutual information bottleneck constraints to disentangle sparse, stable causal features while compressing domain-dependent spurious correlations through variational inference. To address residual spurious correlations, we innovatively design a generative intervention mechanism that breaks local spurious couplings through cross-domain feature recombination while maintaining causal feature semantic consistency via covariance constraints. Furthermore, to mitigate error accumulation in target domain pseudo-labels, we introduce a category-adaptive dynamic calibration strategy, ensuring stable discriminative learning. Extensive experiments on multiple real-world datasets demonstrate that SLOGAN significantly outperforms existing baselines.} }
Endnote
%0 Conference Paper %T Sparse Causal Discovery with Generative Intervention for Unsupervised Graph Domain Adaptation %A Junyu Luo %A Yuhao Tang %A Yiwei Fu %A Xiao Luo %A Zhizhuo Kou %A Zhiping Xiao %A Wei Ju %A Wentao Zhang %A Ming Zhang %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-luo25j %I PMLR %P 41331--41345 %U https://proceedings.mlr.press/v267/luo25j.html %V 267 %X Unsupervised Graph Domain Adaptation (UGDA) leverages labeled source domain graphs to achieve effective performance in unlabeled target domains despite distribution shifts. However, existing methods often yield suboptimal results due to the entanglement of causal-spurious features and the failure of global alignment strategies. We propose SLOGAN (Sparse Causal Discovery with Generative Intervention), a novel approach that achieves stable graph representation transfer through sparse causal modeling and dynamic intervention mechanisms. Specifically, SLOGAN first constructs a sparse causal graph structure, leveraging mutual information bottleneck constraints to disentangle sparse, stable causal features while compressing domain-dependent spurious correlations through variational inference. To address residual spurious correlations, we innovatively design a generative intervention mechanism that breaks local spurious couplings through cross-domain feature recombination while maintaining causal feature semantic consistency via covariance constraints. Furthermore, to mitigate error accumulation in target domain pseudo-labels, we introduce a category-adaptive dynamic calibration strategy, ensuring stable discriminative learning. Extensive experiments on multiple real-world datasets demonstrate that SLOGAN significantly outperforms existing baselines.
APA
Luo, J., Tang, Y., Fu, Y., Luo, X., Kou, Z., Xiao, Z., Ju, W., Zhang, W. & Zhang, M.. (2025). Sparse Causal Discovery with Generative Intervention for Unsupervised Graph Domain Adaptation. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:41331-41345 Available from https://proceedings.mlr.press/v267/luo25j.html.

Related Material