Deep Reasoning Networks for Unsupervised Pattern De-mixing with Constraint Reasoning
D. Chen, Y. Bai, W. Zhao, S. Ament, J. Gregoire, and C. Gomes. Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, page 1500--1509. PMLR, (13--18 Jul 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.
%0 Conference Paper
%1 pmlr-v119-chen20a
%A Chen, Di
%A Bai, Yiwei
%A Zhao, Wenting
%A Ament, Sebastian
%A Gregoire, John
%A Gomes, Carla
%B Proceedings of the 37th International Conference on Machine Learning
%D 2020
%E III, Hal Daumé
%E Singh, Aarti
%I PMLR
%K 2022 ai kde neuro-symbolic seminar
%P 1500--1509
%T Deep Reasoning Networks for Unsupervised Pattern De-mixing with Constraint Reasoning
%U https://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.
@inproceedings{pmlr-v119-chen20a,
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.},
added-at = {2022-07-05T16:11:15.000+0200},
author = {Chen, Di and Bai, Yiwei and Zhao, Wenting and Ament, Sebastian and Gregoire, John and Gomes, Carla},
biburl = {https://www.bibsonomy.org/bibtex/204015e5ca16018b70d65ed44825bb514/soerenmoeller},
booktitle = {Proceedings of the 37th International Conference on Machine Learning},
editor = {III, Hal Daumé and Singh, Aarti},
interhash = {f9d42214520cd328cf2afba4b203aa6f},
intrahash = {04015e5ca16018b70d65ed44825bb514},
keywords = {2022 ai kde neuro-symbolic seminar},
month = {13--18 Jul},
pages = {1500--1509},
pdf = {http://proceedings.mlr.press/v119/chen20a/chen20a.pdf},
publisher = {PMLR},
series = {Proceedings of Machine Learning Research},
timestamp = {2022-07-05T16:11:15.000+0200},
title = {Deep Reasoning Networks for Unsupervised Pattern De-mixing with Constraint Reasoning},
url = {https://proceedings.mlr.press/v119/chen20a.html},
volume = 119,
year = 2020
}