Complexity-Based Methods
作者： Luca Oneto
DOI: 10.1007/978-3-030-24359-3_5
关键词: Empirical error 、 Versa 、 Overfitting 、 Event (probability theory) 、 Large set (Ramsey theory) 、 Computer science 、 Data mining 、 Small set
摘要: The idea behind the complexity-based methods is that if an algorithm chooses from a small set of rules it will probably generalize. Basically, we have and one them has empirical error, risk overfitting data since probability this event happened by chance small. Vice versa large error for high.
springer.com
sci-hub.st 参考文章(90)
Peter L. Bartlett, Olivier Bousquet, Shahar Mendelson, Localized Rademacher Complexities conference on learning theory. pp. 44- 58 ,(2002) , 10.1007/3-540-45435-7_4
Michel Ledoux, Michel Talagrand, Probability in Banach Spaces: Isoperimetry and Processes ,(1991)
Filip M. Mulier, Vladimir Cherkassky, Learning from Data: Concepts, Theory, and Methods John Wiley & Sons, Inc.. ,(1998)
Gilles Blanchard, Pascal Massart, Discussion: Local Rademacher complexities and oracle inequalities in risk minimization Annals of Statistics. ,vol. 34, pp. 2664- 2671 ,(2006) , 10.1214/009053606000001037
Peter L. Bartlett, David S. Rosenberg, The Rademacher Complexity of Co-Regularized Kernel Classes international conference on artificial intelligence and statistics. pp. 396- 403 ,(2007)
Matti Kääriäinen, Generalization error bounds using unlabeled data conference on learning theory. pp. 127- 142 ,(2005) , 10.1007/11503415_9
Dávid Pál, Shai Ben-David, Tyler Lu, Does Unlabeled Data Provably Help? Worst-case Analysis of the Sample Complexity of Semi-Supervised Learning. conference on learning theory. pp. 33- 44 ,(2008)
Xinjia Chen, A Link between Binomial Parameters and Means of Bounded Random Variables arXiv: Probability. ,(2008)
Vladimir Cherkassky, Model complexity control and statisticallearning theory Natural Computing. ,vol. 1, pp. 109- 133 ,(2002) , 10.1023/A:1015007927558
Davide Anguita, Alessandro Ghio, Sandro Ridella, Maximal Discrepancy for Support Vector Machines Neurocomputing. ,vol. 74, pp. 1436- 1443 ,(2011) , 10.1016/J.NEUCOM.2010.12.009