{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,10,11]],"date-time":"2025-10-11T02:17:43Z","timestamp":1760149063247,"version":"build-2065373602"},"reference-count":57,"publisher":"MDPI AG","issue":"7","license":[{"start":{"date-parts":[[2023,7,3]],"date-time":"2023-07-03T00:00:00Z","timestamp":1688342400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000015","name":"US Department of Energy","doi-asserted-by":"publisher","award":["DE-SC0022148"],"award-info":[{"award-number":["DE-SC0022148"]}],"id":[{"id":"10.13039\/100000015","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Symmetry"],"abstract":"<jats:p>A fundamental task in data science is the discovery, description, and identification of any symmetries present in the data. We developed a deep learning methodology for the simultaneous discovery of multiple non-trivial continuous symmetries across an entire labeled dataset. The symmetry transformations and the corresponding generators are modeled with fully connected neural networks trained with a specially constructed loss function, ensuring the desired symmetry properties. The two new elements in this work are the use of a reduced-dimensionality latent space and the generalization to invariant transformations with respect to high-dimensional oracles. The method is demonstrated with several examples on the MNIST digit dataset, where the oracle is provided by the 10-dimensional vector of logits of a trained classifier. We find classes of symmetries that transform each image from the dataset into new synthetic images while conserving the values of the logits. We illustrate these transformations as lines of equal probability (\u201cflows\u201d) in the reduced latent space. These results show that symmetries in the data can be successfully searched for and identified as interpretable non-trivial transformations in the equivalent latent space.<\/jats:p>","DOI":"10.3390\/sym15071352","type":"journal-article","created":{"date-parts":[[2023,7,4]],"date-time":"2023-07-04T01:32:18Z","timestamp":1688434338000},"page":"1352","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":5,"title":["Oracle-Preserving Latent Flows"],"prefix":"10.3390","volume":"15","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-2719-221X","authenticated-orcid":false,"given":"Alexander","family":"Roman","sequence":"first","affiliation":[{"name":"Institute for Fundamental Theory, Physics Department, University of Florida, Gainesville, FL 32611, USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0355-2076","authenticated-orcid":false,"given":"Roy T.","family":"Forestano","sequence":"additional","affiliation":[{"name":"Institute for Fundamental Theory, Physics Department, University of Florida, Gainesville, FL 32611, USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4182-9096","authenticated-orcid":false,"given":"Konstantin T.","family":"Matchev","sequence":"additional","affiliation":[{"name":"Institute for Fundamental Theory, Physics Department, University of Florida, Gainesville, FL 32611, USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3074-998X","authenticated-orcid":false,"given":"Katia","family":"Matcheva","sequence":"additional","affiliation":[{"name":"Institute for Fundamental Theory, Physics Department, University of Florida, Gainesville, FL 32611, USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6683-6463","authenticated-orcid":false,"given":"Eyup B.","family":"Unlu","sequence":"additional","affiliation":[{"name":"Institute for Fundamental Theory, Physics Department, University of Florida, Gainesville, FL 32611, USA"}]}],"member":"1968","published-online":{"date-parts":[[2023,7,3]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"14256","DOI":"10.1073\/pnas.93.25.14256","article-title":"The Role of Symmetry in Fundamental Physics","volume":"93","author":"Gross","year":"1996","journal-title":"Proc. Natl. Acad. Sci. USA"},{"key":"ref_2","first-page":"235","article-title":"Invariante Variationsprobleme","volume":"1918","author":"Noether","year":"1918","journal-title":"Nachrichten Ges. Wiss. G\u00f6ttingen Math. Phys. Kl."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"014","DOI":"10.21468\/SciPostPhys.11.1.014","article-title":"Symmetry meets AI","volume":"11","author":"Barenboim","year":"2021","journal-title":"SciPost Phys."},{"key":"ref_4","unstructured":"Wigner, E., Griffin, J., and Griffin, J. (1959). Group Theory and Its Application to the Quantum Mechanics of Atomic Spectra, Academic Press."},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"010508","DOI":"10.1103\/PhysRevLett.124.010508","article-title":"Discovering Physical Concepts with Neural Networks","volume":"124","author":"Iten","year":"2020","journal-title":"Phys. Rev. Lett."},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"188","DOI":"10.21468\/SciPostPhys.12.6.188","article-title":"Symmetries, safety, and self-supervision","volume":"12","author":"Dillon","year":"2022","journal-title":"SciPost Phys."},{"key":"ref_7","doi-asserted-by":"crossref","first-page":"015010","DOI":"10.1088\/2632-2153\/abbd2d","article-title":"Detecting Symmetries with Neural Networks","volume":"2","author":"Krippendorf","year":"2020","journal-title":"Mach. Learn. Sci. Technol."},{"key":"ref_8","unstructured":"Gruver, N., Finzi, M., Goldblum, M., and Wilson, A.G. (2022). The Lie Derivative for Measuring Learned Equivariance. arXiv."},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"30","DOI":"10.1007\/JHEP07(2022)030","article-title":"An efficient Lorentz equivariant graph neural network for jet tagging","volume":"7","author":"Gong","year":"2022","journal-title":"J. High Energy Phys."},{"key":"ref_10","unstructured":"Li, C., Qu, H., Qian, S., Meng, Q., Gong, S., Zhang, J., Liu, T.Y., and Li, Q. (2022). Does Lorentz-symmetric design boost network performance in jet physics?. arXiv."},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"28","DOI":"10.21468\/SciPostPhys.5.3.028","article-title":"Deep-learned Top Tagging with a Lorentz Layer","volume":"5","author":"Butter","year":"2018","journal-title":"SciPost Phys."},{"key":"ref_12","unstructured":"Bogatskiy, A., Anderson, B., Offermann, J.T., Roussi, M., Miller, D.W., and Kondor, R. (2020). Lorentz Group Equivariant Neural Network for Particle Physics. arXiv."},{"key":"ref_13","doi-asserted-by":"crossref","unstructured":"Hao, Z., Kansal, R., Duarte, J., and Chernyavskaya, N. (2022). Lorentz Group Equivariant Autoencoders. arXiv.","DOI":"10.1140\/epjc\/s10052-023-11633-5"},{"key":"ref_14","doi-asserted-by":"crossref","first-page":"121601","DOI":"10.1103\/PhysRevLett.125.121601","article-title":"Equivariant flow-based sampling for lattice gauge theory","volume":"125","author":"Kanwar","year":"2020","journal-title":"Phys. Rev. Lett."},{"key":"ref_15","unstructured":"Bogatskiy, A., Ganguly, S., Kipf, T., Kondor, R., Miller, D.W., Murnane, D., Offermann, J.T., Pettee, M., Shanahan, P., and Shimmin, C. (2022). Symmetry Group Equivariant Architectures for Physics. arXiv."},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"112008","DOI":"10.1103\/PhysRevD.105.112008","article-title":"Permutationless many-jet event reconstruction with symmetry preserving attention networks","volume":"105","author":"Fenton","year":"2022","journal-title":"Phys. Rev. D"},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"178","DOI":"10.21468\/SciPostPhys.12.5.178","article-title":"SPANet: Generalized permutationless set assignment for particle physics using symmetry preserving attention","volume":"12","author":"Shmakov","year":"2022","journal-title":"SciPost Phys."},{"key":"ref_18","doi-asserted-by":"crossref","first-page":"P08024","DOI":"10.1088\/1748-0221\/17\/08\/P08024","article-title":"A method to challenge symmetries in data with self-supervised learning","volume":"17","author":"Tombs","year":"2022","journal-title":"J. Instrum."},{"key":"ref_19","unstructured":"Lester, C.G., and Tombs, R. (2021). Using unsupervised learning to detect broken symmetries, with relevance to searches for parity violation in nature. (Previously: \u201cStressed GANs snag desserts\u201d). arXiv."},{"key":"ref_20","doi-asserted-by":"crossref","first-page":"508","DOI":"10.1140\/epjc\/s10052-022-10454-2","article-title":"Data-directed search for new physics based on symmetries of the SM","volume":"82","author":"Birman","year":"2022","journal-title":"Eur. Phys. J. C"},{"key":"ref_21","unstructured":"Dersy, A., Schwartz, M.D., and Zhang, X. (2022). Simplifying Polylogarithms with Machine Learning. arXiv."},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"015007","DOI":"10.1088\/2632-2153\/acb2b2","article-title":"SYMBA: Symbolic Computation of Squared Amplitudes in High Energy Physics with Machine Learning","volume":"4","author":"Alnuqaydan","year":"2023","journal-title":"Mach. Learn. Sci. Technol."},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"eaay2631","DOI":"10.1126\/sciadv.aay2631","article-title":"AI Feynman: A Physics-Inspired Method for Symbolic Regression","volume":"6","author":"Udrescu","year":"2020","journal-title":"Sci. Adv."},{"key":"ref_24","unstructured":"Lample, G., and Charton, F. (2019). Deep Learning for Symbolic Mathematics. arXiv."},{"key":"ref_25","unstructured":"d\u2019Ascoli, S., Kamienny, P.A., Lample, G., and Charton, F. (2022). Deep Symbolic Regression for Recurrent Sequences. arXiv."},{"key":"ref_26","unstructured":"Kamienny, P.A., d\u2019Ascoli, S., Lample, G., and Charton, F. (2022). End-to-end symbolic regression with transformers. arXiv."},{"key":"ref_27","unstructured":"Li, J., Yuan, Y., and Shen, H.B. (2022). Symbolic Expression Transformer: A Computer Vision Approach for Symbolic Regression. arXiv."},{"key":"ref_28","unstructured":"Matsubara, Y., Chiba, N., Igarashi, R., Taniai, T., and Ushiku, Y. (2022). Rethinking Symbolic Regression Datasets and Benchmarks for Scientific Discovery. arXiv."},{"key":"ref_29","unstructured":"Cranmer, M.D., Xu, R., Battaglia, P., and Ho, S. (2019). Learning Symbolic Physics with Graph Networks. arXiv."},{"key":"ref_30","first-page":"17429","article-title":"Discovering Symbolic Models from Deep Learning with Inductive Biases","volume":"33","author":"Cranmer","year":"2020","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"ref_31","doi-asserted-by":"crossref","first-page":"2733","DOI":"10.1093\/mnras\/stac1951","article-title":"Modelling the galaxy-halo connection with machine learning","volume":"515","author":"Delgado","year":"2022","journal-title":"Mon. Not. R. Astron. Soc."},{"key":"ref_32","unstructured":"Lemos, P., Jeffrey, N., Cranmer, M., Ho, S., and Battaglia, P. (2022). Rediscovering orbital mechanics with machine learning. arXiv."},{"key":"ref_33","doi-asserted-by":"crossref","first-page":"33","DOI":"10.3847\/1538-4357\/ac610c","article-title":"Analytical Modeling of Exoplanet Transit Spectroscopy with Dimensional Analysis and Symbolic Regression","volume":"930","author":"Matchev","year":"2022","journal-title":"Astrophys. J."},{"key":"ref_34","doi-asserted-by":"crossref","first-page":"110","DOI":"10.1007\/JHEP08(2011)110","article-title":"Construction of a Kinematic Variable Sensitive to the Mass of the Standard Model Higgs Boson in H\u2192WW*\u2192l+\u03bdl\u2212\u03bd\u00af using Symbolic Regression","volume":"8","author":"Choi","year":"2011","journal-title":"J. High Energy Phys."},{"key":"ref_35","unstructured":"Butter, A., Plehn, T., Soybelman, N., and Brehmer, J. (2021). Back to the Formula\u2014LHC Edition. arXiv."},{"key":"ref_36","doi-asserted-by":"crossref","first-page":"055018","DOI":"10.1103\/PhysRevD.107.055018","article-title":"Is the machine smarter than the theorist: Deriving formulas for particle kinematics with symbolic regression","volume":"107","author":"Dong","year":"2023","journal-title":"Phys. Rev. D"},{"key":"ref_37","doi-asserted-by":"crossref","first-page":"793","DOI":"10.1557\/mrc.2019.85","article-title":"Symbolic regression in materials science","volume":"9","author":"Wang","year":"2019","journal-title":"MRS Commun."},{"key":"ref_38","unstructured":"Arechiga, N., Chen, F., Chen, Y.Y., Zhang, Y., Iliev, R., Toyoda, H., and Lyons, K. (2021). Accelerating Understanding of Scientific Experiments with End to End Symbolic Regression. arXiv."},{"key":"ref_39","unstructured":"Cranmer, M., Greydanus, S., Hoyer, S., Battaglia, P., Spergel, D., and Ho, S. (2020). Lagrangian Neural Networks. arXiv."},{"key":"ref_40","doi-asserted-by":"crossref","first-page":"180604","DOI":"10.1103\/PhysRevLett.126.180604","article-title":"Machine Learning Conservation Laws from Trajectories","volume":"126","author":"Liu","year":"2021","journal-title":"Phys. Rev. Lett."},{"key":"ref_41","doi-asserted-by":"crossref","first-page":"033311","DOI":"10.1103\/PhysRevE.100.033311","article-title":"Toward an artificial intelligence physicist for unsupervised learning","volume":"100","author":"Wu","year":"2019","journal-title":"Phys. Rev. E"},{"key":"ref_42","doi-asserted-by":"crossref","first-page":"096030","DOI":"10.1103\/PhysRevD.105.096030","article-title":"Machine learning a manifold","volume":"105","author":"Craven","year":"2022","journal-title":"Phys. Rev. D"},{"key":"ref_43","doi-asserted-by":"crossref","first-page":"033499","DOI":"10.1103\/PhysRevResearch.2.033499","article-title":"Discovering Symmetry Invariants and Conserved Quantities by Interpreting Siamese Neural Networks","volume":"2","author":"Wetzel","year":"2020","journal-title":"Phys. Rev. Res."},{"key":"ref_44","doi-asserted-by":"crossref","first-page":"71","DOI":"10.1142\/S2810939222500058","article-title":"Machine Learning Etudes in Conformal Field Theories","volume":"1","author":"Chen","year":"2023","journal-title":"Int. J. Data Sci. Math. Sci."},{"key":"ref_45","doi-asserted-by":"crossref","first-page":"564","DOI":"10.1016\/j.physletb.2017.10.024","article-title":"Machine-learning the string landscape","volume":"774","author":"He","year":"2017","journal-title":"Phys. Lett. B"},{"key":"ref_46","doi-asserted-by":"crossref","first-page":"157","DOI":"10.1007\/JHEP09(2017)157","article-title":"Machine Learning in the String Landscape","volume":"2017","author":"Carifio","year":"2017","journal-title":"J. High Energy Phys."},{"key":"ref_47","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1016\/j.physrep.2019.09.005","article-title":"Data science applications to string theory","volume":"839","author":"Ruehle","year":"2020","journal-title":"Phys. Rept."},{"key":"ref_48","doi-asserted-by":"crossref","first-page":"096031","DOI":"10.1103\/PhysRevD.105.096031","article-title":"Symmetry discovery with deep learning","volume":"105","author":"Desai","year":"2022","journal-title":"Phys. Rev. D"},{"key":"ref_49","doi-asserted-by":"crossref","first-page":"136297","DOI":"10.1016\/j.physletb.2021.136297","article-title":"Machine learning Lie structures & applications to physics","volume":"817","author":"Chen","year":"2021","journal-title":"Phys. Lett. B"},{"key":"ref_50","doi-asserted-by":"crossref","first-page":"180201","DOI":"10.1103\/PhysRevLett.128.180201","article-title":"Machine Learning Hidden Symmetries","volume":"128","author":"Liu","year":"2022","journal-title":"Phys. Rev. Lett."},{"key":"ref_51","unstructured":"Moskalev, A., Sepliarskaia, A., Sosnovik, I., and Smeulders, A. (2022). LieGG: Studying Learned Lie Group Generators. arXiv."},{"key":"ref_52","doi-asserted-by":"crossref","first-page":"025027","DOI":"10.1088\/2632-2153\/acd989","article-title":"Deep Learning Symmetries and Their Lie Groups, Algebras, and Subalgebras from First Principles","volume":"4","author":"Forestano","year":"2023","journal-title":"Mach. Learn. Sci. Technol."},{"key":"ref_53","doi-asserted-by":"crossref","unstructured":"Forestano, R.T., Matchev, K.T., Matcheva, K., Roman, A., Unlu, E.B., and Verner, S. (2023). Discovering Sparse Representations of Lie Groups with Machine Learning. arXiv.","DOI":"10.1016\/j.physletb.2023.138086"},{"key":"ref_54","unstructured":"Forestano, R.T., Matchev, K.T., Matcheva, K., Roman, A., Unlu, E.B., and Verner, S. (2023, June 02). Oracle Preserving Latent Flows. Available online: https:\/\/github.com\/royforestano\/Deep_Learning_Symmetries\/tree\/main\/Oracle_Preserving_Latent_Flows."},{"key":"ref_55","unstructured":"LeCun, Y., and Cortes, C. (2023, January 05). MNIST Handwritten Digit Database 2010. Available online: https:\/\/keras.io\/api\/datasets\/mnist\/."},{"key":"ref_56","unstructured":"Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., and Antiga, L. (2019). Advances in Neural Information Processing Systems 32, Curran Associates, Inc."},{"key":"ref_57","unstructured":"Goldstein, H. (1980). Classical Mechanics, Addison-Wesley."}],"container-title":["Symmetry"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2073-8994\/15\/7\/1352\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,10]],"date-time":"2025-10-10T20:05:18Z","timestamp":1760126718000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2073-8994\/15\/7\/1352"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,7,3]]},"references-count":57,"journal-issue":{"issue":"7","published-online":{"date-parts":[[2023,7]]}},"alternative-id":["sym15071352"],"URL":"https:\/\/doi.org\/10.3390\/sym15071352","relation":{},"ISSN":["2073-8994"],"issn-type":[{"type":"electronic","value":"2073-8994"}],"subject":[],"published":{"date-parts":[[2023,7,3]]}}}