Adaptive Node Feature Selection for Graph Neural Networks

Madeline Navarro, Ali Azizpour, Santiago Segarra
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:91970-92007, 2026.

Abstract

We propose an adaptive node feature selection approach for graph neural networks (GNNs) that identifies and removes unnecessary features during training. The ability to measure how features contribute to model output is key for interpreting decisions and reducing dimensionality by eliminating unhelpful variables. However, graph-structured data introduces complex dependencies that may be unsuited to classical feature importance metrics. Inspired by this, we present a data-, model-, and task-agnostic method that determines relevant features during training based on changes in validation performance upon permuting feature values. We theoretically motivate our approach by characterizing how the relationships between node data and graph structure influences GNN performance. Empirically, we show that (i) our highly general approach rivals the performance of tailored feature selection approaches that exploit prior assumptions; (ii) we return meaningful feature importance scores well before the GNN is fully trained; and (iii) our scores demonstrably extract relevant properties that inform feature importance for various graph learning settings.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-navarro26a, title = {Adaptive Node Feature Selection for Graph Neural Networks}, author = {Navarro, Madeline and Azizpour, Ali and Segarra, Santiago}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {91970--92007}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/navarro26a/navarro26a.pdf}, url = {https://proceedings.mlr.press/v306/navarro26a.html}, abstract = {We propose an adaptive node feature selection approach for graph neural networks (GNNs) that identifies and removes unnecessary features during training. The ability to measure how features contribute to model output is key for interpreting decisions and reducing dimensionality by eliminating unhelpful variables. However, graph-structured data introduces complex dependencies that may be unsuited to classical feature importance metrics. Inspired by this, we present a data-, model-, and task-agnostic method that determines relevant features during training based on changes in validation performance upon permuting feature values. We theoretically motivate our approach by characterizing how the relationships between node data and graph structure influences GNN performance. Empirically, we show that (i) our highly general approach rivals the performance of tailored feature selection approaches that exploit prior assumptions; (ii) we return meaningful feature importance scores well before the GNN is fully trained; and (iii) our scores demonstrably extract relevant properties that inform feature importance for various graph learning settings.} }
Endnote
%0 Conference Paper %T Adaptive Node Feature Selection for Graph Neural Networks %A Madeline Navarro %A Ali Azizpour %A Santiago Segarra %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-navarro26a %I PMLR %P 91970--92007 %U https://proceedings.mlr.press/v306/navarro26a.html %V 306 %X We propose an adaptive node feature selection approach for graph neural networks (GNNs) that identifies and removes unnecessary features during training. The ability to measure how features contribute to model output is key for interpreting decisions and reducing dimensionality by eliminating unhelpful variables. However, graph-structured data introduces complex dependencies that may be unsuited to classical feature importance metrics. Inspired by this, we present a data-, model-, and task-agnostic method that determines relevant features during training based on changes in validation performance upon permuting feature values. We theoretically motivate our approach by characterizing how the relationships between node data and graph structure influences GNN performance. Empirically, we show that (i) our highly general approach rivals the performance of tailored feature selection approaches that exploit prior assumptions; (ii) we return meaningful feature importance scores well before the GNN is fully trained; and (iii) our scores demonstrably extract relevant properties that inform feature importance for various graph learning settings.
APA
Navarro, M., Azizpour, A. & Segarra, S.. (2026). Adaptive Node Feature Selection for Graph Neural Networks. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:91970-92007 Available from https://proceedings.mlr.press/v306/navarro26a.html.

Related Material