A Probabilistic Circuit Framework for Interpretable Graph PU Learning

Sagad Hamid, Dooho Lee, Myeong Kong, Tanya Braun, Jaemin Yoo
Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, PMLR 337:1913-1930, 2026.

Abstract

How can we make graph positive-unlabeled ({PU}) learning interpretable? Existing methods jointly process node features and edge information, which obscures their interaction and makes predictions difficult to interpret. In this paper, we propose a novel interpretable graph {PU} learning framework that explicitly decouples feature and edge processing, enabling multi-level interpretability. Our framework first produces a core prediction from node features using probabilistic circuits (PCs) and then refines it using edge information, providing graph-level interpretability by exposing how graph structure affects predictions. For feature-based prediction, we construct two PCs through a careful split of the training nodes, yielding node-level interpretability by highlighting which nodes support the separation of positive and negative instances. Finally, by leveraging the tractability of PCs, we obtain feature-level interpretability via feature attribute marginalization, which quantifies attribute impact and importance while revealing interactions and dependencies. Experiments on 10 datasets show that our framework achieves strong performance while substantially improving interpretability. The code is available at https://github.com/hagad1/graphpu-cpu.

Cite this Paper

BibTeX
@InProceedings{pmlr-v337-hamid26a, title = {A Probabilistic Circuit Framework for Interpretable Graph {PU} Learning}, author = {Hamid, Sagad and Lee, Dooho and Kong, Myeong and Braun, Tanya and Yoo, Jaemin}, booktitle = {Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence}, pages = {1913--1930}, year = {2026}, editor = {Perković, Emilija and Malinsky, Daniel}, volume = {337}, series = {Proceedings of Machine Learning Research}, month = {17--21 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v337/main/assets/hamid26a/hamid26a.pdf}, url = {https://proceedings.mlr.press/v337/hamid26a.html}, abstract = {How can we make graph positive-unlabeled ({PU}) learning interpretable? Existing methods jointly process node features and edge information, which obscures their interaction and makes predictions difficult to interpret. In this paper, we propose a novel interpretable graph {PU} learning framework that explicitly decouples feature and edge processing, enabling multi-level interpretability. Our framework first produces a core prediction from node features using probabilistic circuits (PCs) and then refines it using edge information, providing graph-level interpretability by exposing how graph structure affects predictions. For feature-based prediction, we construct two PCs through a careful split of the training nodes, yielding node-level interpretability by highlighting which nodes support the separation of positive and negative instances. Finally, by leveraging the tractability of PCs, we obtain feature-level interpretability via feature attribute marginalization, which quantifies attribute impact and importance while revealing interactions and dependencies. Experiments on 10 datasets show that our framework achieves strong performance while substantially improving interpretability. The code is available at https://github.com/hagad1/graphpu-cpu.} }
Endnote
%0 Conference Paper %T A Probabilistic Circuit Framework for Interpretable Graph PU Learning %A Sagad Hamid %A Dooho Lee %A Myeong Kong %A Tanya Braun %A Jaemin Yoo %B Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Emilija Perković %E Daniel Malinsky %F pmlr-v337-hamid26a %I PMLR %P 1913--1930 %U https://proceedings.mlr.press/v337/hamid26a.html %V 337 %X How can we make graph positive-unlabeled ({PU}) learning interpretable? Existing methods jointly process node features and edge information, which obscures their interaction and makes predictions difficult to interpret. In this paper, we propose a novel interpretable graph {PU} learning framework that explicitly decouples feature and edge processing, enabling multi-level interpretability. Our framework first produces a core prediction from node features using probabilistic circuits (PCs) and then refines it using edge information, providing graph-level interpretability by exposing how graph structure affects predictions. For feature-based prediction, we construct two PCs through a careful split of the training nodes, yielding node-level interpretability by highlighting which nodes support the separation of positive and negative instances. Finally, by leveraging the tractability of PCs, we obtain feature-level interpretability via feature attribute marginalization, which quantifies attribute impact and importance while revealing interactions and dependencies. Experiments on 10 datasets show that our framework achieves strong performance while substantially improving interpretability. The code is available at https://github.com/hagad1/graphpu-cpu.
APA
Hamid, S., Lee, D., Kong, M., Braun, T. & Yoo, J.. (2026). A Probabilistic Circuit Framework for Interpretable Graph PU Learning. Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 337:1913-1930 Available from https://proceedings.mlr.press/v337/hamid26a.html.

Related Material