Neurosymbolic models based on hybrids of convolutional neural networks and decision trees

Rasul Kairgeldin, Miguel Á. Carreira-Perpiñán
Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, PMLR 284:796-813, 2025.

Abstract

Building on previous work, we propose a specific form of neurosymbolic model consisting of the composition of convolutional neural network layers with a sparse oblique classification tree (having hyperplane splits using few features). This can be seen as a neural feature extraction that finds a more suitable representation of the input space followed by a form of rule-based reasoning to arrive at a decision that can be explained. We show how to control the sparsity across the different decision nodes of the tree and its effect on the explanations produced. We demonstrate this on image classification tasks and show, among other things, that relatively small subsets of neurons are entirely responsible for the classification into specific classes, and that the neurons’ receptive fields focus on areas of the image that provide best discrimination.

Cite this Paper

BibTeX
@InProceedings{pmlr-v284-kairgeldin25a, title = {Neurosymbolic models based on hybrids of convolutional neural networks and decision trees}, author = {Kairgeldin, Rasul and Carreira-Perpi\~{n}\'{a}n, Miguel \'{A}.}, booktitle = {Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning}, pages = {796--813}, year = {2025}, editor = {H. Gilpin, Leilani and Giunchiglia, Eleonora and Hitzler, Pascal and van Krieken, Emile}, volume = {284}, series = {Proceedings of Machine Learning Research}, month = {08--10 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v284/main/assets/kairgeldin25a/kairgeldin25a.pdf}, url = {https://proceedings.mlr.press/v284/kairgeldin25a.html}, abstract = {Building on previous work, we propose a specific form of neurosymbolic model consisting of the composition of convolutional neural network layers with a sparse oblique classification tree (having hyperplane splits using few features). This can be seen as a neural feature extraction that finds a more suitable representation of the input space followed by a form of rule-based reasoning to arrive at a decision that can be explained. We show how to control the sparsity across the different decision nodes of the tree and its effect on the explanations produced. We demonstrate this on image classification tasks and show, among other things, that relatively small subsets of neurons are entirely responsible for the classification into specific classes, and that the neurons’ receptive fields focus on areas of the image that provide best discrimination.} }
Endnote
%0 Conference Paper %T Neurosymbolic models based on hybrids of convolutional neural networks and decision trees %A Rasul Kairgeldin %A Miguel Á. Carreira-Perpiñán %B Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning %C Proceedings of Machine Learning Research %D 2025 %E Leilani H. Gilpin %E Eleonora Giunchiglia %E Pascal Hitzler %E Emile van Krieken %F pmlr-v284-kairgeldin25a %I PMLR %P 796--813 %U https://proceedings.mlr.press/v284/kairgeldin25a.html %V 284 %X Building on previous work, we propose a specific form of neurosymbolic model consisting of the composition of convolutional neural network layers with a sparse oblique classification tree (having hyperplane splits using few features). This can be seen as a neural feature extraction that finds a more suitable representation of the input space followed by a form of rule-based reasoning to arrive at a decision that can be explained. We show how to control the sparsity across the different decision nodes of the tree and its effect on the explanations produced. We demonstrate this on image classification tasks and show, among other things, that relatively small subsets of neurons are entirely responsible for the classification into specific classes, and that the neurons’ receptive fields focus on areas of the image that provide best discrimination.
APA
Kairgeldin, R. & Carreira-Perpiñán, M.Á.. (2025). Neurosymbolic models based on hybrids of convolutional neural networks and decision trees. Proceedings of The 19th International Conference on Neurosymbolic Learning and Reasoning, in Proceedings of Machine Learning Research 284:796-813 Available from https://proceedings.mlr.press/v284/kairgeldin25a.html.

Related Material