Neural Processing Letters

, Volume 50, Issue 2, pp 1845–1860 | Cite as

Robust Class-Specific Autoencoder for Data Cleaning and Classification in the Presence of Label Noise

  • Weining Zhang
  • Dong Wang
  • Xiaoyang TanEmail author


We present a simple but effective method for data cleaning and classification in the presence of label noise. The fundamental idea is to treat the data points with label noise as outliers of the class indicated by the corresponding noisy label. This essentially allows us to deal with the traditional supervised problem of classification with label noise as an unsupervised one, i.e., identifying outliers from each class. However, finding such dubious observations (outliers) from each class is challenging in general. We therefore propose to reduce their potential influence using class-specific feature learning by autoencoder. Particularly, we learn for each class a feature space using all the samples labeled as that class, including those with noisy (but unknown to us) labels. Furthermore, in order to solve the situation when the noise is relatively high, we propose a weighted class-specific autoencoder by considering the effect of each data point on the postulated model. To fully exploit the advantage of the learned class-specific feature space, we use a minimum reconstruction error based method for finding out the outliers (label noise) and solving the classification task. Experiments on several datasets show that the proposed method achieves state of the art performance on the task of data cleaning and classification with noisy labels.


Class-specific autoencoder Label noise Classification Data cleaning Outliers 



The authors thank the anonymous reviewers for their valuable comments and suggestions. This work is partially supported by National Science Foundation of China (61672280, 61373060, 61732006), AI+ Project of NUAA (56XZA18009), Jiangsu 333 Project (BRA2017377) and Qing Lan Project.


  1. 1.
    Abellán J, Masegosa AR (2010) Bagging decision trees on data sets with classification noise. In: International symposium on foundations of information and knowledge systems. Springer, pp 248–265Google Scholar
  2. 2.
    Aggarwal CC (ed) (2015) Outlier analysis. In: Data mining. Springer, Berlin, pp 237–263Google Scholar
  3. 3.
    Bengio Y, Lamblin P, Popovici D, Larochelle H (2007) Greedy layer-wise training of deep networks. In: Advances in neural information processing systems, pp 153–160Google Scholar
  4. 4.
    Biggio B, Nelson B, Laskov P (2011) Support vector machines under adversarial label noise. ACML 20:97–112Google Scholar
  5. 5.
    Brodley CE, Friedl MA (1999) Identifying mislabeled training data. J Artif Intell Res 11:131–167CrossRefGoogle Scholar
  6. 6.
    Chandola V, Banerjee A, Kumar V (2009) Anomaly detection: a survey. ACM Comput Surv (CSUR) 41(3):15CrossRefGoogle Scholar
  7. 7.
    Ekambaram R, Fefilatyev S, Shreve M, Kramer K, Hall LO, Goldgof DB, Kasturi R (2016) Active cleaning of label noise. Pattern Recognit 51:463–480CrossRefGoogle Scholar
  8. 8.
    Fefilatyev S, Shreve M, Kramer K, Hall L, Goldgof D, Kasturi R, Daly K, Remsen A, Bunke H (2012) Label-noise reduction with support vector machines. In: 2012 21st International Conference on Pattern Recognition (ICPR). IEEE, pp 3504–3508Google Scholar
  9. 9.
    Frénay B, Verleysen M (2014) Classification in the presence of label noise: a survey. IEEE Trans Neural Netw Learn Syst 25(5):845–869CrossRefGoogle Scholar
  10. 10.
    Gupta K, Majumdar A (2017) Imposing class-wise feature similarity in stacked autoencoders by nuclear norm regularization. Neural Process Lett 48:1–15Google Scholar
  11. 11.
    Hawkins DM (1980) Identification of outliers, vol 11. Springer, BerlinCrossRefGoogle Scholar
  12. 12.
    Hoz EDL, Hoz EDL, Ortiz A, Ortega J, Martnez-lvarez A (2014) Feature selection by multi-objective optimisation: application to network anomaly detection by hierarchical self-organising maps. Knowl Based Syst 71:322–338CrossRefGoogle Scholar
  13. 13.
    Huber PJ (2011) Robust statistics. Springer, BerlinGoogle Scholar
  14. 14.
    Ipeirotis PG, Provost F, Wang J (2010) Quality management on amazon mechanical turk. In: Proceedings of the ACM SIGKDD workshop on human computation. ACM, pp 64–67Google Scholar
  15. 15.
    Jeatrakul P, Wong KW, Fung CC (2010) Data cleaning for classification using misclassification analysis. J Adv Comput Intell Intell Inform 14(3):297–302CrossRefGoogle Scholar
  16. 16.
    Kamimura R, Nakanishi S (1995) Feature detectors by autoencoders: decomposition of input patterns into atomic features by neural networks. Neural Process Lett 2(6):17–22CrossRefGoogle Scholar
  17. 17.
    Kingma DP, Welling M (2013) Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114
  18. 18.
    Krishna RA, Hata K, Chen S, Kravitz J, Shamma DA, Fei-Fei L, Bernstein MS (2016) Embracing error to enable rapid crowdsourcing. In: Proceedings of the 2016 CHI conference on human factors in computing systems. ACM, pp 3167–3179Google Scholar
  19. 19.
    Lab R, Gunnar Rtsch PD (2001) Soft margins for adaboost. Mach Learn 42(3):287–320CrossRefGoogle Scholar
  20. 20.
    LeCun Y, Bottou L, Bengio Y, Haffner P (1998) Gradient-based learning applied to document recognition. Proc IEEE 86(11):2278–2324CrossRefGoogle Scholar
  21. 21.
    Li W, Wang L, Li W, Agustsson E, Van Gool L (2017) Webvision database: visual learning and understanding from web data. arXiv preprint arXiv:1708.02862
  22. 22.
    Liu T, Tao D (2016) Classification with noisy labels by importance reweighting. IEEE Trans Pattern Anal Mach Intell 38(3):447–461CrossRefGoogle Scholar
  23. 23.
    Makhzani A, Frey B (2013) K-sparse autoencoders. arXiv preprint arXiv:1312.5663
  24. 24.
    Maria J, Amaro J, Falcao G, Alexandre LA (2016) Stacked autoencoders using low-power accelerated architectures for object recognition in autonomous systems. Neural Process Lett 43(2):445–458CrossRefGoogle Scholar
  25. 25.
    Natarajan N, Dhillon IS, Ravikumar PK, Tewari A (2013) Learning with noisy labels. In: Advances in neural information processing systems, pp 1196–1204Google Scholar
  26. 26.
    Nettleton DF, Orriols-Puig A, Fornells A (2010) A study of the effect of different types of noise on the precision of supervised learning techniques. Artif Intell Rev 33(4):275–306CrossRefGoogle Scholar
  27. 27.
    Pechenizkiy M, Tsymbal A, Puuronen S, Pechenizkiy O (2006) Class noise and supervised learning in medical domains: the effect of feature extraction. In: 19th IEEE international symposium on computer-based medical systems. CBMS 2006. IEEE, pp 708–713Google Scholar
  28. 28.
    Pruengkarn R, Wong KW, Fung CC (2016) Data cleaning using complementary fuzzy support vector machine technique. In: International conference on neural information processing. Springer, pp 160–167Google Scholar
  29. 29.
    Qian Q, Hu J, Jin R, Pei J, Zhu S (2014) Distance metric learning using dropout: a structured regularization approach. In: Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, pp 323–332Google Scholar
  30. 30.
    Rebbapragada UD (2010) Strategic targeting of outliers for expert review. Ph.D. thesis, Tufts UniversityGoogle Scholar
  31. 31.
    Rifai S, Vincent P, Muller X, Glorot X, Bengio Y (2011) Contractive auto-encoders: explicit invariance during feature extraction. In: Proceedings of the 28th international conference on machine learning (ICML-11), pp 833–840Google Scholar
  32. 32.
    Rolnick D, Veit A, Belongie S, Shavit N (2017) Deep learning is robust to massive label noise. arXiv preprint arXiv:1705.10694
  33. 33.
    Rtsch G, Schlkopf B, Smola AJ, Mika S, Onoda T, Mller KR (2000) Robust ensemble learning for data mining. In: Pacific-Asia conference on knowledge discovery and data mining, Current Issues and New Applications, pp 341–344Google Scholar
  34. 34.
    Sáez JA, Galar M, Luengo J, Herrera F (2014) Analyzing the presence of noise in multi-class problems: alleviating its influence with the one-vs-one decomposition. Knowl Inf Syst 38(1):179–206CrossRefGoogle Scholar
  35. 35.
    Teng CM (2005) Dealing with data corruption in remote sensing. In: International conference on advances in intelligent data analysis, pp 452–463Google Scholar
  36. 36.
    Vidal R, Ma Y, Sastry S (2005) Generalized principal component analysis (GPCA). IEEE Trans Pattern Anal Mach Intell 27(12):1945–1959CrossRefGoogle Scholar
  37. 37.
    Vidal R, Ma Y, Sastry SS (2016) Robust principal component analysis. In: Antman SS (ed) Generalized Principal Component Analysis. Springer, Berlin pp 63–122Google Scholar
  38. 38.
    Vincent P, Larochelle H, Bengio Y, Manzagol PA (2008) Extracting and composing robust features with denoising autoencoders. In: Proceedings of the 25th international conference on machine learning. ACM, pp 1096–1103Google Scholar
  39. 39.
    Vincent P, Larochelle H, Lajoie I, Bengio Y, Manzagol PA (2010) Stacked denoising autoencoders: learning useful representations in a deep network with a local denoising criterion. J Mach Learn Res 11:3371–3408MathSciNetzbMATHGoogle Scholar
  40. 40.
    Wang D, Tan X (2014) Robust distance metric learning in the presence of label noise. In: AAAI, pp 1321–1327Google Scholar
  41. 41.
    Wang H, Nie F, Huang H (2014) Robust distance metric learning via simultaneous l1-norm minimization and maximization. In: International conference on machine learning, pp 1836–1844Google Scholar
  42. 42.
    Yang L, Jin R, Sukthankar R (2012) Bayesian active distance metric learning. arXiv preprint arXiv:1206.5283
  43. 43.
    Yang T, Mahdavi M, Jin R, Zhang L, Zhou Y (2012) Multiple kernel learning from noisy labels by stochastic programming. arXiv preprint arXiv:1206.4629
  44. 44.
    Zhang W, Rekaya R, Bertrand K (2005) A method for predicting disease subtypes in presence of misclassification among training samples using gene expression: application to human breast cancer. Bioinformatics 22(3):317–325CrossRefGoogle Scholar
  45. 45.
    Zhang W, Wang D, Tan X (2018) Data cleaning and classification in the presence of label noise with class-specific autoencoder. In: International symposium on neural networksGoogle Scholar
  46. 46.
    Zhu X, Wu X (2004) Class noise vs. attribute noise: a quantitative study. Artif Intell Rev 22(3):177–210Google Scholar

Copyright information

© Springer Science+Business Media, LLC, part of Springer Nature 2018

Authors and Affiliations

  1. 1.College of Computer Science and TechnologyNanjing University of Aeronautics and AstronauticsNanjingChina
  2. 2.Collaborative Innovation Center of Novel Software Technology and IndustrializationNanjingChina

Personalised recommendations