Deep Reasoning Networks for Unsupervised Pattern De-mixing with Constraint Reasoning

Di Chen, Yiwei Bai, Wenting Zhao, Sebastian Ament, John Gregoire, Carla Gomes
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:1500-1509, 2020.

Abstract

We introduce Deep Reasoning Networks (DRNets), an end-to-end framework that combines deep learning with constraint reasoning for solving pattern de-mixing problems, typically in an unsupervised or very-weakly-supervised setting. DRNets exploit problem structure and prior knowledge by tightly combining constraint reasoning with stochastic-gradient-based neural network optimization. Our motivating task is from materials discovery and concerns inferring crystal structures of materials from X-ray diffraction data (Crystal-Structure-Phase-Mapping). Given the complexity of its underlying scientific domain, we start by introducing DRNets on an analogous but much simpler task: de-mixing overlapping hand-written Sudokus (Multi-MNIST-Sudoku). On Multi-MNIST-Sudoku, DRNets almost perfectly recovered the mixed Sudokus’ digits, with 100% digit accuracy, outperforming the supervised state-of-the-art MNIST de-mixing models. On Crystal-Structure-Phase-Mapping, DRNets significantly outperform the state of the art and experts’ capabilities, recovering more precise and physically meaningful crystal structures.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-chen20a, title = {Deep Reasoning Networks for Unsupervised Pattern De-mixing with Constraint Reasoning}, author = {Chen, Di and Bai, Yiwei and Zhao, Wenting and Ament, Sebastian and Gregoire, John and Gomes, Carla}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {1500--1509}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/chen20a/chen20a.pdf}, url = {http://proceedings.mlr.press/v119/chen20a.html}, abstract = {We introduce Deep Reasoning Networks (DRNets), an end-to-end framework that combines deep learning with constraint reasoning for solving pattern de-mixing problems, typically in an unsupervised or very-weakly-supervised setting. DRNets exploit problem structure and prior knowledge by tightly combining constraint reasoning with stochastic-gradient-based neural network optimization. Our motivating task is from materials discovery and concerns inferring crystal structures of materials from X-ray diffraction data (Crystal-Structure-Phase-Mapping). Given the complexity of its underlying scientific domain, we start by introducing DRNets on an analogous but much simpler task: de-mixing overlapping hand-written Sudokus (Multi-MNIST-Sudoku). On Multi-MNIST-Sudoku, DRNets almost perfectly recovered the mixed Sudokus’ digits, with 100% digit accuracy, outperforming the supervised state-of-the-art MNIST de-mixing models. On Crystal-Structure-Phase-Mapping, DRNets significantly outperform the state of the art and experts’ capabilities, recovering more precise and physically meaningful crystal structures.} }
Endnote
%0 Conference Paper %T Deep Reasoning Networks for Unsupervised Pattern De-mixing with Constraint Reasoning %A Di Chen %A Yiwei Bai %A Wenting Zhao %A Sebastian Ament %A John Gregoire %A Carla Gomes %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-chen20a %I PMLR %P 1500--1509 %U http://proceedings.mlr.press/v119/chen20a.html %V 119 %X We introduce Deep Reasoning Networks (DRNets), an end-to-end framework that combines deep learning with constraint reasoning for solving pattern de-mixing problems, typically in an unsupervised or very-weakly-supervised setting. DRNets exploit problem structure and prior knowledge by tightly combining constraint reasoning with stochastic-gradient-based neural network optimization. Our motivating task is from materials discovery and concerns inferring crystal structures of materials from X-ray diffraction data (Crystal-Structure-Phase-Mapping). Given the complexity of its underlying scientific domain, we start by introducing DRNets on an analogous but much simpler task: de-mixing overlapping hand-written Sudokus (Multi-MNIST-Sudoku). On Multi-MNIST-Sudoku, DRNets almost perfectly recovered the mixed Sudokus’ digits, with 100% digit accuracy, outperforming the supervised state-of-the-art MNIST de-mixing models. On Crystal-Structure-Phase-Mapping, DRNets significantly outperform the state of the art and experts’ capabilities, recovering more precise and physically meaningful crystal structures.
APA
Chen, D., Bai, Y., Zhao, W., Ament, S., Gregoire, J. & Gomes, C.. (2020). Deep Reasoning Networks for Unsupervised Pattern De-mixing with Constraint Reasoning. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:1500-1509 Available from http://proceedings.mlr.press/v119/chen20a.html.

Related Material