publication . Preprint . 2017

Learning with Confident Examples: Rank Pruning for Robust Classification with Noisy Labels

Northcutt, Curtis G.; Wu, Tailin; Chuang, Isaac L.;
Open Access English
  • Published: 04 May 2017
Abstract
Noisy PN learning is the problem of binary classification when training examples may be mislabeled (flipped) uniformly with noise rate rho1 for positive examples and rho0 for negative examples. We propose Rank Pruning (RP) to solve noisy PN learning and the open problem of estimating the noise rates, i.e. the fraction of wrong positive and negative labels. Unlike prior solutions, RP is time-efficient and general, requiring O(T) for any unrestricted choice of probabilistic classifier with T fitting time. We prove RP has consistent noise estimation and equivalent expected risk as learning with uncorrupted labels in ideal conditions, and derive closed-form solution...
Subjects
free text keywords: Statistics - Machine Learning, Computer Science - Learning
Download from
38 references, page 1 of 3

Aha, D. W., Kibler, D., and Albert, M. K. Instance-based learning algorithms. Mach. Learn., 6(1):37-66, 1991.

Angelova, A., Abu-Mostafam, Y., and Perona, P. Pruning training sets for learning of object categories. In CVPR, volume 1, pp. 494-501. IEEE, 2005. [OpenAIRE]

Angluin, D. and Laird, P. Learning from noisy examples. Machine Learning, 2(4):343-370, 1988. [OpenAIRE]

Blanchard, G., Lee, G., and Scott, C. Semi-supervised novelty detection. J. Mach. Learn. Res., 11:2973-3009, December 2010. ISSN 1532-4435.

Blum, A. and Mitchell, T. Combining labeled and unlabeled data with co-training. In 11th Conf. on COLT, pp. 92-100, New York, NY, USA, 1998. ACM.

Blum, M., Floyd, R. W., Pratt, V., Rivest, R. L., and Tarjan, R. E. Time bounds for selection. J. Comput. Syst. Sci., 7 (4):448-461, August 1973. ISSN 0022-0000. [OpenAIRE]

Breiman, L. Bagging predictors. Machine Learning, 24(2):123-140, August 1996. ISSN 0885-6125.

Chapelle, O. and Vapnik, V. Model selection for support vector machines. In Proc. of 12th NIPS, pp. 230-236, Cambridge, MA, USA, 1999.

Chollet, F. Keras CIFAR CNN, 2016a. bit.ly/2mVKR3d.

Chollet, F. Keras MNIST CNN, 2016b. bit.ly/2nKiqJv.

Claesen, M., Smet, F. D., Suykens, J. A., and Moor, B. D. A robust ensemble approach to learn from positive and unlabeled data using fSVMg base models. Neurocomputing, 160:73 - 84, 2015. ISSN 0925-2312. [OpenAIRE]

Davis, J. and Goadrich, M. The relationship between precision-recall and roc curves. In Proc. of 23rd ICML, pp. 233-240, NYC, NY, USA, 2006. ACM.

Elkan, C. and Noto, K. Learning classifiers from only positive and unlabeled data. In Proc. of 14th KDD, pp. 213-220, NYC, NY, USA, 2008. ACM. [OpenAIRE]

Hempstalk, K., Frank, E., and Witten, I. H. One-class classification by combining density and class probability estimation. In Proc. of ECML-PKDD, pp. 505-519, Berlin, Heidelberg, 2008. Springer-Verlag. [OpenAIRE]

Krizhevsky, A., Nair, V., and Hinton, G. Cifar-10 (canadian institute for advanced research).

38 references, page 1 of 3
Abstract
Noisy PN learning is the problem of binary classification when training examples may be mislabeled (flipped) uniformly with noise rate rho1 for positive examples and rho0 for negative examples. We propose Rank Pruning (RP) to solve noisy PN learning and the open problem of estimating the noise rates, i.e. the fraction of wrong positive and negative labels. Unlike prior solutions, RP is time-efficient and general, requiring O(T) for any unrestricted choice of probabilistic classifier with T fitting time. We prove RP has consistent noise estimation and equivalent expected risk as learning with uncorrupted labels in ideal conditions, and derive closed-form solution...
Subjects
free text keywords: Statistics - Machine Learning, Computer Science - Learning
Download from
38 references, page 1 of 3

Aha, D. W., Kibler, D., and Albert, M. K. Instance-based learning algorithms. Mach. Learn., 6(1):37-66, 1991.

Angelova, A., Abu-Mostafam, Y., and Perona, P. Pruning training sets for learning of object categories. In CVPR, volume 1, pp. 494-501. IEEE, 2005. [OpenAIRE]

Angluin, D. and Laird, P. Learning from noisy examples. Machine Learning, 2(4):343-370, 1988. [OpenAIRE]

Blanchard, G., Lee, G., and Scott, C. Semi-supervised novelty detection. J. Mach. Learn. Res., 11:2973-3009, December 2010. ISSN 1532-4435.

Blum, A. and Mitchell, T. Combining labeled and unlabeled data with co-training. In 11th Conf. on COLT, pp. 92-100, New York, NY, USA, 1998. ACM.

Blum, M., Floyd, R. W., Pratt, V., Rivest, R. L., and Tarjan, R. E. Time bounds for selection. J. Comput. Syst. Sci., 7 (4):448-461, August 1973. ISSN 0022-0000. [OpenAIRE]

Breiman, L. Bagging predictors. Machine Learning, 24(2):123-140, August 1996. ISSN 0885-6125.

Chapelle, O. and Vapnik, V. Model selection for support vector machines. In Proc. of 12th NIPS, pp. 230-236, Cambridge, MA, USA, 1999.

Chollet, F. Keras CIFAR CNN, 2016a. bit.ly/2mVKR3d.

Chollet, F. Keras MNIST CNN, 2016b. bit.ly/2nKiqJv.

Claesen, M., Smet, F. D., Suykens, J. A., and Moor, B. D. A robust ensemble approach to learn from positive and unlabeled data using fSVMg base models. Neurocomputing, 160:73 - 84, 2015. ISSN 0925-2312. [OpenAIRE]

Davis, J. and Goadrich, M. The relationship between precision-recall and roc curves. In Proc. of 23rd ICML, pp. 233-240, NYC, NY, USA, 2006. ACM.

Elkan, C. and Noto, K. Learning classifiers from only positive and unlabeled data. In Proc. of 14th KDD, pp. 213-220, NYC, NY, USA, 2008. ACM. [OpenAIRE]

Hempstalk, K., Frank, E., and Witten, I. H. One-class classification by combining density and class probability estimation. In Proc. of ECML-PKDD, pp. 505-519, Berlin, Heidelberg, 2008. Springer-Verlag. [OpenAIRE]

Krizhevsky, A., Nair, V., and Hinton, G. Cifar-10 (canadian institute for advanced research).

38 references, page 1 of 3
Powered by OpenAIRE Open Research Graph
Any information missing or wrong?Report an Issue