Skip to main content
Erschienen in: Journal of Applied Mathematics and Computing 2/2023

11.10.2022 | Original Research

A modified stochastic quasi-Newton algorithm for summing functions problem in machine learning

verfasst von: Xiaoxuan Chen, Haishan Feng

Erschienen in: Journal of Applied Mathematics and Computing | Ausgabe 2/2023

Einloggen

Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.

search-config
loading …

Abstract

In this paper, a new stochastic quasi-Newton method (SQN) is proposed which has a different approximation of the Hessian inverse matrix \(H_k\). The modified quasi-Newton Broyden–Fletcher–Goldfarb–Shanno (BFGS) formula which has a better approximation to Hessian matrix has not only the gradient variation but also the function value. Because of the special nature of the sum function, the mini-batch setting is built in the algorithm, and less compution cost can be guaranteed. The number of iterations reduce to at most \(O(\varepsilon ^{-\frac{1}{1-\beta }})\). The convergence analysis is established in this paper. The numerical experiments show that this algorithm is competitive to other algorithms.

Sie haben noch keine Lizenz? Dann Informieren Sie sich jetzt über unsere Produkte:

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

  • über 102.000 Bücher
  • über 537 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Maschinenbau + Werkstoffe
  • Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

Springer Professional "Wirtschaft"

Online-Abonnement

Mit Springer Professional "Wirtschaft" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 340 Zeitschriften

aus folgenden Fachgebieten:

  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Versicherung + Risiko




Jetzt Wissensvorsprung sichern!

Springer Professional "Technik"

Online-Abonnement

Mit Springer Professional "Technik" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 390 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Maschinenbau + Werkstoffe




 

Jetzt Wissensvorsprung sichern!

Literatur
1.
Zurück zum Zitat Bottou, L.: Large-scale machine learning with stochastic gradient descent. In: Proceedings of COMPSTAT’, pp. 177–186 (2010) Bottou, L.: Large-scale machine learning with stochastic gradient descent. In: Proceedings of COMPSTAT’, pp. 177–186 (2010)
2.
Zurück zum Zitat Byrd, R.H., Hansen, S.L., Nocedal, J., et al.: A stochastic quasi-Newton method for large-scale optimization. SIAM J. Optim. 26, 1008–1031 (2012)MathSciNetCrossRefMATH Byrd, R.H., Hansen, S.L., Nocedal, J., et al.: A stochastic quasi-Newton method for large-scale optimization. SIAM J. Optim. 26, 1008–1031 (2012)MathSciNetCrossRefMATH
3.
Zurück zum Zitat Cotter, A., Shamir, O., Srebro, N., Sridharan, K.: Better mini-batch algorithms via accelerated gradient methods. In: Proceedings of the Advances in Neural Information Processing Systems, pp. 1647–1655 (2011) Cotter, A., Shamir, O., Srebro, N., Sridharan, K.: Better mini-batch algorithms via accelerated gradient methods. In: Proceedings of the Advances in Neural Information Processing Systems, pp. 1647–1655 (2011)
4.
Zurück zum Zitat Durrett, R.: Probability: Theory and Examples. Cambridge University Press, London (2010)CrossRefMATH Durrett, R.: Probability: Theory and Examples. Cambridge University Press, London (2010)CrossRefMATH
5.
6.
Zurück zum Zitat Ghadimi, S., Lan, G.: Stochastic first-and zeroth-order methods for nonconvex stochastic programming. SIAM J. Optim. 23, 2341–2368 (2013)MathSciNetCrossRefMATH Ghadimi, S., Lan, G.: Stochastic first-and zeroth-order methods for nonconvex stochastic programming. SIAM J. Optim. 23, 2341–2368 (2013)MathSciNetCrossRefMATH
7.
Zurück zum Zitat Hassan, B., Mohammed, T.: A new variants of quasi-newton equation based on the quadratic function for unconstrained optimization. Indones. J. Electr. Eng. Comput. Sci. 2, 701–708 (2020) Hassan, B., Mohammed, T.: A new variants of quasi-newton equation based on the quadratic function for unconstrained optimization. Indones. J. Electr. Eng. Comput. Sci. 2, 701–708 (2020)
8.
Zurück zum Zitat Lan, G.: An optimal method for stochastic convex optimization. Technical report, Georgia Institute of Technology (2009) Lan, G.: An optimal method for stochastic convex optimization. Technical report, Georgia Institute of Technology (2009)
9.
Zurück zum Zitat Li, M., Zhang, T., Chen, Y., et al.: Efficient mini-batch training for stochastic optimization. In: Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 661–670 (2014) Li, M., Zhang, T., Chen, Y., et al.: Efficient mini-batch training for stochastic optimization. In: Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 661–670 (2014)
10.
Zurück zum Zitat Mason, L., Baxter, J., Bartlett, P., Frean, M.: Boosting algorithms as gradient descent in function space. Adv. Neural Inf. Process. Syst. 12, 512–518 (1999) Mason, L., Baxter, J., Bartlett, P., Frean, M.: Boosting algorithms as gradient descent in function space. Adv. Neural Inf. Process. Syst. 12, 512–518 (1999)
11.
Zurück zum Zitat Mokhtari, A., Eisen, M., Ribeiro, A.: IQN: An incremental quasi-Newton method with local superlinear convergence rate. SIAM J. Optim. 28, 1670–1698 (2018)MathSciNetCrossRefMATH Mokhtari, A., Eisen, M., Ribeiro, A.: IQN: An incremental quasi-Newton method with local superlinear convergence rate. SIAM J. Optim. 28, 1670–1698 (2018)MathSciNetCrossRefMATH
12.
Zurück zum Zitat Nemirovski, A., Juditsky, A., Lan, G., Shapiro, A.: Robust stochastic approximation approach to stochastic programming. SIAM J. Optim. 19(4), 1574–1609 (2009)MathSciNetCrossRefMATH Nemirovski, A., Juditsky, A., Lan, G., Shapiro, A.: Robust stochastic approximation approach to stochastic programming. SIAM J. Optim. 19(4), 1574–1609 (2009)MathSciNetCrossRefMATH
13.
Zurück zum Zitat Powell, M.J.D.: Some global convergence properties of a variable metric algorithm for minimization without exact line searches. In: Cottle, R.W., Lemke, C.E. (eds.) Nonlinear Programming, Philadelphia (1976) Powell, M.J.D.: Some global convergence properties of a variable metric algorithm for minimization without exact line searches. In: Cottle, R.W., Lemke, C.E. (eds.) Nonlinear Programming, Philadelphia (1976)
15.
Zurück zum Zitat Schmidt, M., Roux, N., Bach, F.: Minimizing finite sums with the stochastic average gr1adient. Math. Program. 162, 83–112 (2017)MathSciNetCrossRefMATH Schmidt, M., Roux, N., Bach, F.: Minimizing finite sums with the stochastic average gr1adient. Math. Program. 162, 83–112 (2017)MathSciNetCrossRefMATH
16.
Zurück zum Zitat Shang, F., Zhou, K., Liu, H., et al.: VR-SGD: a simple stochastic variance reduction method for machine learning. IEEE Trans. Knowl. Data Eng. 32, 188–202 (2018)CrossRef Shang, F., Zhou, K., Liu, H., et al.: VR-SGD: a simple stochastic variance reduction method for machine learning. IEEE Trans. Knowl. Data Eng. 32, 188–202 (2018)CrossRef
17.
Zurück zum Zitat Wang, X., Ma, S., Goldfarb, D., Liu, W.: Stochastic quasi-Newton methods for nonconvex stochastic optimization. SIAM J. Optim. 27, 927–956 (2017)MathSciNetCrossRefMATH Wang, X., Ma, S., Goldfarb, D., Liu, W.: Stochastic quasi-Newton methods for nonconvex stochastic optimization. SIAM J. Optim. 27, 927–956 (2017)MathSciNetCrossRefMATH
18.
Zurück zum Zitat Wei, Z., Yu, G., Yuan, G., et al.: The superlinear convergence of a modified BFGS-type method for unconstrained optimization. Comput. Optim. Appl. 29, 315–332 (2004)MathSciNetCrossRefMATH Wei, Z., Yu, G., Yuan, G., et al.: The superlinear convergence of a modified BFGS-type method for unconstrained optimization. Comput. Optim. Appl. 29, 315–332 (2004)MathSciNetCrossRefMATH
19.
Zurück zum Zitat Wei, Z., Yu, G., Yuan, G., Lian, Z.: The superlinear convergence of a modified BFGS-type method for unconstrained optimization. Comput. Optim. Appl. 29, 315–332 (2004)MathSciNetCrossRefMATH Wei, Z., Yu, G., Yuan, G., Lian, Z.: The superlinear convergence of a modified BFGS-type method for unconstrained optimization. Comput. Optim. Appl. 29, 315–332 (2004)MathSciNetCrossRefMATH
20.
Zurück zum Zitat Yang, Z., Wang, C., Zang, Y., et al.: Mini-batch algorithms with Barzilai–Borwein update step. Neurocomputing 314, 177–185 (2018)CrossRef Yang, Z., Wang, C., Zang, Y., et al.: Mini-batch algorithms with Barzilai–Borwein update step. Neurocomputing 314, 177–185 (2018)CrossRef
21.
Zurück zum Zitat Yuan, G., Wei, Z.: Convergence analysis of a modified BFGS method on convex minimizations. Comput. Optim. Appl. 47, 237–255 (2010)MathSciNetCrossRefMATH Yuan, G., Wei, Z.: Convergence analysis of a modified BFGS method on convex minimizations. Comput. Optim. Appl. 47, 237–255 (2010)MathSciNetCrossRefMATH
22.
Zurück zum Zitat Yuan, G., Zhou, Y., Wang, L., et al.: Stochastic bigger subspace algorithms for nonconvex stochastic optimization. IEEE Access 9, 119818–119829 (2021)CrossRef Yuan, G., Zhou, Y., Wang, L., et al.: Stochastic bigger subspace algorithms for nonconvex stochastic optimization. IEEE Access 9, 119818–119829 (2021)CrossRef
23.
Zurück zum Zitat Zhang, J.Z., Deng, N.Y., Chen, L.H.: New quasi-Newton equation and related methods for unconstrained optimization. J. Optim. Theory Appl. 102, 147–167 (1999)MathSciNetCrossRefMATH Zhang, J.Z., Deng, N.Y., Chen, L.H.: New quasi-Newton equation and related methods for unconstrained optimization. J. Optim. Theory Appl. 102, 147–167 (1999)MathSciNetCrossRefMATH
Metadaten
Titel
A modified stochastic quasi-Newton algorithm for summing functions problem in machine learning
verfasst von
Xiaoxuan Chen
Haishan Feng
Publikationsdatum
11.10.2022
Verlag
Springer Berlin Heidelberg
Erschienen in
Journal of Applied Mathematics and Computing / Ausgabe 2/2023
Print ISSN: 1598-5865
Elektronische ISSN: 1865-2085
DOI
https://doi.org/10.1007/s12190-022-01800-4

Weitere Artikel der Ausgabe 2/2023

Journal of Applied Mathematics and Computing 2/2023 Zur Ausgabe

Premium Partner