IDEAS home Printed from https://ideas.repec.org/p/tse/wpaper/125195.html
   My bibliography  Save this paper

A mathematical model for automatic differentiation in machine learning

Author

Listed:
  • Bolte, Jérôme
  • Pauwels, Edouard

Abstract

Automatic differentiation, as implemented today, does not have a simple mathematical model adapted to the needs of modern machine learning. In this work we articulate the relationships between differentiation of programs as implemented in practice and differentiation of nonsmooth functions. To this end we provide a simple class of functions, a nonsmooth calculus, and show how they apply to stochastic approximation methods. We also evidence the issue of artificial critical points created by algorithmic differentiation and show how usual methods avoid these points with probability one.

Suggested Citation

  • Bolte, Jérôme & Pauwels, Edouard, 2021. "A mathematical model for automatic differentiation in machine learning," TSE Working Papers 21-1184, Toulouse School of Economics (TSE).
    • Handle: RePEc:tse:wpaper:125195
    as

    Download full text from publisher

    File URL: https://www.tse-fr.eu/sites/default/files/TSE/documents/doc/wp/2021/wp_tse_1184.pdf
    File Function: Full Text
    Download Restriction: no
    ---><---

    References listed on IDEAS

    as
    1. Bolte, Jérôme & Castera, Camille & Pauwels, Edouard & Févotte, Cédric, 2019. "An Inertial Newton Algorithm for Deep Learning," TSE Working Papers 19-1043, Toulouse School of Economics (TSE).
    2. Michel Benaim & Josef Hofbauer & Sylvain Sorin, 2005. "Stochastic Approximations and Differential Inclusions II: Applications," Levine's Bibliography 784828000000000098, UCLA Department of Economics.
    3. Michel Benaïm & Josef Hofbauer & Sylvain Sorin, 2005. "Stochastic Approximations and Differential Inclusions; Part II: Applications," Working Papers hal-00242974, HAL.
    4. Bolte, Jérôme & Pauwels, Edouard, 2019. "Conservative set valued fields, automatic differentiation, stochastic gradient methods and deep learning," TSE Working Papers 19-1044, Toulouse School of Economics (TSE).
    Full references (including those not matched with items on IDEAS)

    Citations

    Citations are extracted by the CitEc Project, subscribe to its RSS feed for this item.
    as


    Cited by:

    1. Edouard Pauwels, 2021. "Incremental Without Replacement Sampling in Nonconvex Optimization," Journal of Optimization Theory and Applications, Springer, vol. 190(1), pages 274-299, July.

    Most related items

    These are the items that most often cite the same works as this one and are cited by the same works as this one.
    1. Bolte, Jérôme & Le, Tam & Pauwels, Edouard & Silveti-Falls, Antonio, 2022. "Nonsmooth Implicit Differentiation for Machine Learning and Optimization," TSE Working Papers 22-1314, Toulouse School of Economics (TSE).
    2. Bolte, Jérôme & Pauwels, Edouard, 2019. "Conservative set valued fields, automatic differentiation, stochastic gradient methods and deep learning," TSE Working Papers 19-1044, Toulouse School of Economics (TSE).
    3. Benaïm, Michel & Hofbauer, Josef & Hopkins, Ed, 2009. "Learning in games with unstable equilibria," Journal of Economic Theory, Elsevier, vol. 144(4), pages 1694-1709, July.
    4. Akimoto, Youhei & Auger, Anne & Hansen, Nikolaus, 2022. "An ODE method to prove the geometric convergence of adaptive stochastic algorithms," Stochastic Processes and their Applications, Elsevier, vol. 145(C), pages 269-307.
    5. Michel Benaïm & Josef Hofbauer & Sylvain Sorin, 2006. "Stochastic Approximations and Differential Inclusions, Part II: Applications," Mathematics of Operations Research, INFORMS, vol. 31(4), pages 673-695, November.
    6. Vinayaka G. Yaji & Shalabh Bhatnagar, 2020. "Stochastic Recursive Inclusions in Two Timescales with Nonadditive Iterate-Dependent Markov Noise," Mathematics of Operations Research, INFORMS, vol. 45(4), pages 1405-1444, November.
    7. Swenson, Brian & Murray, Ryan & Kar, Soummya, 2020. "Regular potential games," Games and Economic Behavior, Elsevier, vol. 124(C), pages 432-453.
    8. Cason, Timothy N. & Friedman, Daniel & Hopkins, Ed, 2010. "Testing the TASP: An experimental investigation of learning in games with unstable equilibria," Journal of Economic Theory, Elsevier, vol. 145(6), pages 2309-2331, November.
    9. Bervoets, Sebastian & Faure, Mathieu, 2020. "Convergence in games with continua of equilibria," Journal of Mathematical Economics, Elsevier, vol. 90(C), pages 25-30.
    10. Bervoets, Sebastian & Faure, Mathieu, 2019. "Stability in games with continua of equilibria," Journal of Economic Theory, Elsevier, vol. 179(C), pages 131-162.
    11. Andrés Contreras & Juan Peypouquet, 2019. "Asymptotic Equivalence of Evolution Equations Governed by Cocoercive Operators and Their Forward Discretizations," Journal of Optimization Theory and Applications, Springer, vol. 182(1), pages 30-48, July.
    12. Michel Benaïm & Mathieu Faure, 2013. "Consistency of Vanishingly Smooth Fictitious Play," Mathematics of Operations Research, INFORMS, vol. 38(3), pages 437-450, August.
    13. Michel Benaim & Olivier Raimond, 2007. "Simulated Annealing, Vertex-Reinforced Random Walks and Learning in Games," Levine's Bibliography 122247000000001702, UCLA Department of Economics.
    14. Josef Hofbauer & Sylvain Sorin & Yannick Viossat, 2009. "Time Average Replicator and Best-Reply Dynamics," Mathematics of Operations Research, INFORMS, vol. 34(2), pages 263-269, May.
    15. Esponda, Ignacio & Pouzo, Demian & Yamamoto, Yuichi, 2021. "Asymptotic behavior of Bayesian learners with misspecified models," Journal of Economic Theory, Elsevier, vol. 195(C).
    16. Ignacio Esponda & Demian Pouzo & Yuichi Yamamoto, 2019. "Asymptotic Behavior of Bayesian Learners with Misspecified Models," Papers 1904.08551, arXiv.org, revised Oct 2019.
    17. Arunselvan Ramaswamy & Shalabh Bhatnagar, 2017. "A Generalization of the Borkar-Meyn Theorem for Stochastic Recursive Inclusions," Mathematics of Operations Research, INFORMS, vol. 42(3), pages 648-661, August.
    18. Candogan, Ozan & Ozdaglar, Asuman & Parrilo, Pablo A., 2013. "Dynamics in near-potential games," Games and Economic Behavior, Elsevier, vol. 82(C), pages 66-90.
    19. Sandholm, William H., 2015. "Population Games and Deterministic Evolutionary Dynamics," Handbook of Game Theory with Economic Applications,, Elsevier.
    20. Viossat, Yannick & Zapechelnyuk, Andriy, 2013. "No-regret dynamics and fictitious play," Journal of Economic Theory, Elsevier, vol. 148(2), pages 825-842.

    More about this item

    NEP fields

    This paper has been announced in the following NEP Reports:

    Statistics

    Access and download statistics

    Corrections

    All material on this site has been provided by the respective publishers and authors. You can help correct errors and omissions. When requesting a correction, please mention this item's handle: RePEc:tse:wpaper:125195. See general information about how to correct material in RePEc.

    If you have authored this item and are not yet registered with RePEc, we encourage you to do it here. This allows to link your profile to this item. It also allows you to accept potential citations to this item that we are uncertain about.

    If CitEc recognized a bibliographic reference but did not link an item in RePEc to it, you can help with this form .

    If you know of missing items citing this one, you can help us creating those links by adding the relevant references in the same way as above, for each refering item. If you are a registered author of this item, you may also want to check the "citations" tab in your RePEc Author Service profile, as there may be some citations waiting for confirmation.

    For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: the person in charge (email available below). General contact details of provider: https://edirc.repec.org/data/tsetofr.html .

    Please note that corrections may take a couple of weeks to filter through the various RePEc services.

    IDEAS is a RePEc service. RePEc uses bibliographic data supplied by the respective publishers.