Abstract
The paper presents an estimation of overfitting probability for VKF-method of algebraic machine learning in the simplest case of Boolean algebra without counter-examples. The model uses the Vapnik—Chervonenkis proposal to minimize the empirical risk. Asymptotically the probability of overfitting errors for a fixed fraction of test examples tends to zero faster than exponentially decrease if the description length and the number of requested hypotheses go to infinity.
Similar content being viewed by others
REFERENCES
Vinogradov, D.V., Machine learning based on similarity operation, Commun. Comput. Inf. Sci., 2018, vol. 934, pp. 46–59.
Vinogradov, D.V., Probabilistic-combinatorial formal learning method based on theory of grids, Doctoral (Phys.-Math.) Dissertation, Moscow: Federal Research Center Computer Science and Control, Russian Academy of Sciences, 2018.
Vinogradov, D.V., The rate of convergence to the limit of the probability of encountering an accidental similarity in the presence of counter examples, Autom. Doc. Math. Linguist., 2018, vol. 52, no. 1, pp. 35–37. https://doi.org/10.3103/S0005105518010090
Vorontsov, K.V., Combinatorial approach to estimation of quality of learnable algorithms, Matematicheskie voprosy kibernetiki (Mathematical Questions of Cybernetics), Moscow: Fizmatlit, 2004, pp. 5–36.
Vapnik, V.N. and Chervonenkis, A.Ya., Teoriya raspoznavaniya obrazov (Pattern Recognition Theory), Moscow: Nauka, 1974.
Feller, W., An Introduction to Probability Theory and Its Applications, Chapman & Hall, 1958, 2nd ed., vol. 1.
DSM-metod avtomaticheskogo porozhdeniya gipotez: logicheskie i epistemologicheskie osnovaniya (JSM Method of Automated Hypothesis Generation: Logical and Epistemological Foundations), Anshakov, O.M., Ed., Moscow: URSS, 2009.
ACKNOWLEDGMENTS
The author is grateful to his colleagues at the Dorodnicyn Computing Center, Federal Research Center Computer Science and Control, Russian Academy of Sciences, for support and useful discussions. Special thanks to L.A. Yakimova for collaboration, discussion, and support.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
The author declares that he has no conflicts of interest.
About this article
Cite this article
Vinogradov, D. Estimation of Overfitting Degree of Algebraic Machine Learning in Boolean Algebra. Autom. Doc. Math. Linguist. 56, 160–162 (2022). https://doi.org/10.3103/S0005105522030098
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.3103/S0005105522030098