Abstract
Experimental data bases are typically very large and high dimensional. To learn from them requires to recognize important features (a pattern), often present at scales different to that of the recorded data. Following the experience collected in statistical mechanics and thermodynamics, the process of recognizing the pattern (the learning process) can be seen as a dissipative time evolution driven by entropy from a detailed level of description to less detailed. This is the way thermodynamics enters machine learning. On the other hand, reversible (typically Hamiltonian) evolution is propagation within the levels of description, that is also to be recognized. This is how Poisson geometry enters machine learning. Learning to handle free surface liquids and damped rigid body rotation serves as an illustration.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
In this work we mean pattern recognition in a broad sense as a process of extracting any information from the data. The dissipative-driven pattern recognition can be then imagined as a retouche of the original data leading to recognition of the important aspects.
- 2.
The Poisson bracket corresponding to the Poisson bivector is \(\{F,G\}=\langle F_x| {^\uparrow \mathbf {L}}| G_x \rangle \), where \(\langle \bullet |\bullet \rangle \) denotes a scalar product.
- 3.
The more detailed level is referred to as the upper while the less detailed (reduced) as lower.
- 4.
- 5.
It is often assumed that the reduced manifold keeps the structure of a cotangent bundle, such that a reversible evolution is generated by the canonical Poisson bivector (equipped with entropy) as on the original manifold. Therefore, the reduced dynamics can be interpreted as dynamics of a lower number of (quasi-)particles, since otherwise an another Poisson bivector would have to be sought. This is not, however, strictly necessary nor a limitation of the method, see for instance [49, 50].
- 6.
Corresponding to dissipation potential \(\Xi =\frac{1}{2} y^*_a {{^\downarrow }M}^{ab} y^*_b\).
References
Gesamtausgabe, L.: Ludwig Boltzmann Gesamtausgabe - Collected Works (1983)
Gorban, A.N., Grechuk, B., Tyukin, I.Y.: Augmented artificial intelligence: a conceptual framework (2018)
Kosambi, D.D.: J. Indian Math. Soc. 7, 76 (1943)
Golub, G., Van Loan, C.: Matrix Computations. Johns Hopkins Studies in the Mathematical Sciences. Johns Hopkins University Press (2013). https://books.google.cz/books?id=X5YfsuCWpxMC
Roweis, S.T., Saul, L.K.: Science 290(5500), 2323 (2000)
Wasserman, L.: Ann. Rev. Stat. Appl. 5(1), 501 (2018)
Brunton, S., Proctor, J., Kutz, J.: Proceedings of the National Academy of Sciences (2016). https://doi.org/10.1073/pnas.1517384113
Kaiser, E., Kutz, J., Brunton, S.: Discovering conservation laws from data for control (2018)
Kevrekidis, Y., Samaey, G.: Scholarpedia 5(9), 4847 (2010)
Weinan, E., Commun. Math. Stat. 5, 1 (2017). https://doi.org/10.1007/s40304-017-0103-z
Weinan, E., Han, J., Li, Q.: Res. Math. Sci. 6(1), 10 (2018). https://doi.org/10.1007/s40687-018-0172-y
Moya, B., Gonzalez, D., Alfaro, I., Chinesta, F., Cueto, E.: Comput. Mech. 64(2), 511 (2019). https://doi.org/10.1007/s00466-019-01705-3
Maes, C., Netočný, K.: eprint arXiv:cond-mat/0202501 (2002)
Grmela, M., Öttinger, H.C.: Phys. Rev. E 56, 6620 (1997). https://doi.org/10.1103/PhysRevE.56.6620
Öttinger, H.C., Grmela, M.: Phys. Rev. E 56, 6633 (1997). https://doi.org/10.1103/PhysRevE.56.6633
Öttinger, H.: Beyond Equilibrium Thermodynamics. Wiley (2005)
Pavelka, M., Klika, V., Grmela, M.: Multiscale Thermo-Dynamics (De Gruyter, Berlin. Boston (2018). https://doi.org/10.1515/9783110350951. http://www.degruyter.com/view/books/9783110350951/9783110350951/9783110350951.xml
Pavelka, M., Klika, V., Grmela, M.: Phys. Rev. E 90 (2014)
Onsager, L.: Phys. Rev. 37, 405 (1931). https://doi.org/10.1103/PhysRev.37.405. http://link.aps.org/doi/10.1103/PhysRev.37.405
Onsager, L.: Phys. Rev. 38, 2265 (1931). https://doi.org/10.1103/PhysRev.38.2265
Casimir, H.B.G.: Rev. Mod. Phys. 17, 343 (1945). https://doi.org/10.1103/RevModPhys.17.343
de Groot, S.R., Mazur, P.: Non-equilibrium Thermodynamics. Dover Publications, New York (1984)
Grmela, M., Klika, V., Pavelka, M.: Phys. Rev. E 92 (2015)
Grmela, M.: J. Stat. Phys. 166(2), 282 (2017)
Grmela, M.: J. Phys. Commun 2 (2018)
Shannon, C.E.: Bell Syst. Tech. J. 27, 379 (1948)
Jaynes, E.T.: Phys. Rev. 106(4), 620 (1957)
Gorban, A., Karlin, I.: Invariant Manifolds for Physical and Chemical Kinetics. Lecture Notes in Physics. Springer (2005). http://books.google.cz/books?id=hjvjPmL5rPwC
Klika, V., Pavelka, M., Vágner, P., Grmela, M.: Entropy 21, 715 (2019). https://doi.org/10.3390/e21070715
Chapman, S., Cowling, T., Burnett, D., Cercignani, C.: The Mathematical Theory of Non-uniform Gases: An Account of the Kinetic Theory of Viscosity, Thermal Conduction and Diffusion in Gases. Cambridge Mathematical Library. Cambridge University Press (1990). https://books.google.cz/books?id=Cbp5JP2OTrwC
Callen, H.: Thermodynamics: An Introduction to the Physical Theories of Equilibrium Thermostatics and Irreversible Thermodynamics. Wiley (1960). http://books.google.cz/books?id=mf5QAAAAMAAJ
Turkington, B.: J. Stat. Phys. 152, 569 (2013)
Pavelka, M., Klika, V., Grmela, M.: J. Stat. Phys. Accepted (2020)
Ehrenfest, P., Ehrenfest, T.: The Conceptual Foundations of the Statistical Approach in Mechanics. Dover Books on Physics. Dover Publications (1990)
Gorban, A.N., Karlin, I.V., Öttinger, H.C., Tatarinova, L.L.: Phys. Rev. E 63 (2001)
Karlin, I.V., Tatarinova, L.L., Gorban, A.N., Öttinger, H.C.: Physica A: Stat. Mech. Appl. 327(3–4), 399 (2003)
Pavelka, M., Klika, V., Grmela, M.: Entropy 20 (2018)
Pavelka, M., Klika, V., Grmela, M.: Physica D: Nonlinear Phenomena 399, 193 (2019). https://doi.org/10.1016/j.physd.2019.06.006. http://www.sciencedirect.com/science/article/pii/S0167278918305232
Grmela, M.: Comput. Math. Appl. 65(10), 1457 (2013). https://doi.org/10.1016/j.camwa.2012.11.019. http://www.sciencedirect.com/science/article/pii/S0898122112006803
Grmela, M.: Entropy 16(3), 1652 (2014). https://doi.org/10.3390/e16031652
Čapek, K.l.: R.U.R. (Rossum’s Universal Robots). Penguin Books, London (2004)
Chatterjee, A.: Curr. Sci. 78, 7 (2000)
Brockett, R.: Linear Algebra Appl. 146, 79 (1991). https://doi.org/10.1016/0024-3795(91)90021-N
Bloch,A.M., Krishnaprasad, P.S., Marsden, J.E., Ratiu, T.S.: Annales de l’I.H.P. Analyse non linéaire 11(1), 37 (1994)
Absil, A.: Int. J. Unconvent. Comput. 2(4), 291 (2006)
Gingold, R., Monaghan, J.: Mon. Not. R. Astron. Soc. 181(3), 375 (1977)
Español, P., Revenga, M.: Phys. Rev. E 67 (2003). https://doi.org/10.1103/PhysRevE.67.026705. https://link.aps.org/doi/10.1103/PhysRevE.67.026705
Ellero, M., Espaňol, P.: Appl. Math. Mech. 39(1), 103 (2018)
González, D., Chinesta, F., Cueto, E.: Continuum Mechanics and Thermodynamics (2018). https://doi.org/10.1007/s00161-018-0677-z
González, D., Chinesta, F., Cueto, E.: Front. Mat. 6, 14 (2019)
Brockett, R.: Linear Algebra Appl. 122-124, 761 (1989). Special Issue on Linear Systems and Control
Landau, L., Lifshitz, E.: Mechanics. Butterworth-Heinemann (1976)
Arnold, V.: Annales de l’institut Fourier 16(1), 319 (1966)
Marsden, J., Ratiu, T., Weinstein, A.: Trans. Am. Math. Soc. 281(1), 147 (1984). https://doi.org/10.2307/1999527
Simo, J.C., Marsden, J.E., Krishnaprasad, P.S.: Arch. Rat. Mech. Anal. 104(2), 125 (1988). https://doi.org/10.1007/BF00251673. https://doi.org/10.1007/BF00251673
Pavelka, M., Šípka, M.: Machine-learning-rigid-body (2020). https://github.com/enaipi/machine-learning-rigid-body.git
Acknowledgements
We are grateful to Václav Klika for discussing the manuscript.
F.Ch. thanks ESI Group through its research chair at “Arts et Métiers ParisTech”, whose first invited position was Prof. M. Grmela, for performing the researches here addressed. F. Ch. also knowledges Dr. Alain de Rouvray by the rich and inspiring discussions on pattern recognition as the first step towards machine learning and artificial intelligence, motivating the preset work. The support from ANR (Agence Nationale de la Recherche, France) through its grant AAPG2018 DataBEST is also gratefully acknowledged.
E.C. also acknowledges the financial support of ESI Group through the project “Simulated Reality”. The support given by the Spanish Ministry of Economy and Competitiveness through Grant number DPI2017-85139-C2-1-R, and by the Regional Government of Aragon and the European Social Fund, research group T88, is also greatly acknowledged.
M.G. was supported by the Natural Sciences and Engineering Research Council of Canada, Grants 3100319 and 3100735.
B.M. acknowledges the support of the Spanish Ministry of Science, Innovation and Universities through grant number PRE2018-083211.
M.P. and M.Š were supported by Czech Science Foundation, Project No. 20-22092S. M.P. was supported by Charles University Research Program No. UNCE/SCI/023.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Chinesta, F., Cueto, E., Grmela, M., Moya, B., Pavelka, M., Šípka, M. (2021). Learning Physics from Data: A Thermodynamic Interpretation. In: Barbaresco, F., Nielsen, F. (eds) Geometric Structures of Statistical Physics, Information Geometry, and Learning. SPIGL 2020. Springer Proceedings in Mathematics & Statistics, vol 361. Springer, Cham. https://doi.org/10.1007/978-3-030-77957-3_14
Download citation
DOI: https://doi.org/10.1007/978-3-030-77957-3_14
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-77956-6
Online ISBN: 978-3-030-77957-3
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)