Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity

Ryan Henderson, Djork-Arné Clevert, Floriane Montanari
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:4203-4213, 2021.

Abstract

Rationalizing which parts of a molecule drive the predictions of a molecular graph convolutional neural network (GCNN) can be difficult. To help, we propose two simple regularization techniques to apply during the training of GCNNs: Batch Representation Orthonormalization (BRO) and Gini regularization. BRO, inspired by molecular orbital theory, encourages graph convolution operations to generate orthonormal node embeddings. Gini regularization is applied to the weights of the output layer and constrains the number of dimensions the model can use to make predictions. We show that Gini and BRO regularization can improve the accuracy of state-of-the-art GCNN attribution methods on artificial benchmark datasets. In a real-world setting, we demonstrate that medicinal chemists significantly prefer explanations extracted from regularized models. While we only study these regularizers in the context of GCNNs, both can be applied to other types of neural networks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-henderson21a, title = {Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity}, author = {Henderson, Ryan and Clevert, Djork-Arn{\'e} and Montanari, Floriane}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {4203--4213}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/henderson21a/henderson21a.pdf}, url = {https://proceedings.mlr.press/v139/henderson21a.html}, abstract = {Rationalizing which parts of a molecule drive the predictions of a molecular graph convolutional neural network (GCNN) can be difficult. To help, we propose two simple regularization techniques to apply during the training of GCNNs: Batch Representation Orthonormalization (BRO) and Gini regularization. BRO, inspired by molecular orbital theory, encourages graph convolution operations to generate orthonormal node embeddings. Gini regularization is applied to the weights of the output layer and constrains the number of dimensions the model can use to make predictions. We show that Gini and BRO regularization can improve the accuracy of state-of-the-art GCNN attribution methods on artificial benchmark datasets. In a real-world setting, we demonstrate that medicinal chemists significantly prefer explanations extracted from regularized models. While we only study these regularizers in the context of GCNNs, both can be applied to other types of neural networks.} }
Endnote
%0 Conference Paper %T Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity %A Ryan Henderson %A Djork-Arné Clevert %A Floriane Montanari %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-henderson21a %I PMLR %P 4203--4213 %U https://proceedings.mlr.press/v139/henderson21a.html %V 139 %X Rationalizing which parts of a molecule drive the predictions of a molecular graph convolutional neural network (GCNN) can be difficult. To help, we propose two simple regularization techniques to apply during the training of GCNNs: Batch Representation Orthonormalization (BRO) and Gini regularization. BRO, inspired by molecular orbital theory, encourages graph convolution operations to generate orthonormal node embeddings. Gini regularization is applied to the weights of the output layer and constrains the number of dimensions the model can use to make predictions. We show that Gini and BRO regularization can improve the accuracy of state-of-the-art GCNN attribution methods on artificial benchmark datasets. In a real-world setting, we demonstrate that medicinal chemists significantly prefer explanations extracted from regularized models. While we only study these regularizers in the context of GCNNs, both can be applied to other types of neural networks.
APA
Henderson, R., Clevert, D. & Montanari, F.. (2021). Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:4203-4213 Available from https://proceedings.mlr.press/v139/henderson21a.html.

Related Material