An empirical analysis of the effect of Label Noise on Artificial Neural Networks learning

• Lívia de Azevedo Federal Rural University of Rio de Janeiro
• Filipe Braida Federal Rural University of Rio de Janeiro

DOI: https://doi.org/10.5753/sbbd_estendido.2022.21837

Abstract

Label noise consists of the class labeling error. It can negatively affect the performance of a model, and it can vary with respect to the model chosen. For this reason, works have emerged that evaluate the natural resistance of Machine Learning models to label noise. Therefore, it would be relevant to evaluate the natural resistance of the artificial neural network to label noise, given its relevance to Deep Learning. The purpose of this work is to perform an experiment to evaluate the influence of label noise on artificial neural networks, training them on noisy databases. The results showed that the network complexity can influence their resistance to label noise.

Keywords: Label Noise, Artificial Neural Networks, Machine Learning

References

Algan, G. and Ulusoy, I. (2020). Label noise types and their effects on deep learning. arXiv preprint arXiv:2003.10471.

Frénay, B. and Verleysen, M. (2013). Classification in the presence of label noise: a survey. IEEE transactions on neural networks and learning systems, 25(5):845-869.

Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep Learning. MIT Press. https://www.deeplearningbook.org/.

Han, B., Yao, Q., Liu, T., Niu, G., Tsang, I. W., Kwok, J. T., and Sugiyama, M. (2020). A survey of label-noise representation learning: Past, present and future. arXiv preprint arXiv:2011.04406.

LeCun, Y. (1998). The mnist database of handwritten digits. http://yann.lecun.com/exdb/mnist/.

Nettleton, D. F., Orriols-Puig, A., and Fornells, A. (2010). A study of the effect of different types of noise on the precision of supervised learning techniques. Artificial intelligence review, 33(4):275-306.

Patrini, G., Rozza, A., Krishna Menon, A., Nock, R., and Qu, L. (2017). Making deep neural networks robust to label noise: A loss correction approach. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1944-1952.

Rolnick, D., Veit, A., Belongie, S., and Shavit, N. (2017). Deep learning is robust to massive label noise. arXiv preprint arXiv:1705.10694.

Rusiecki, A. (2020). Standard dropout as remedy for training deep neural networks with label noise. In International Conference on Dependability and Complex Systems, pages 534-542. Springer.

Russell, S. J. and Norvig, P. (2021). Artificial Intelligence: A Modern Approach. Pearson, global edition.

Simard, P. Y., Steinkraus, D., Platt, J. C., et al. (2003). Best practices for convolutional neural networks applied to visual document analysis. In Icdar, volume 3.

Song, H., Kim, M., Park, D., Shin, Y., and Lee, J.-G. (2020). Learning from noisy labels with deep neural networks: A survey. arXiv preprint arXiv:2007.08199.

Xiao, H., Rasul, K., and Vollgraf, R. (2017). Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms. arXiv preprint arXiv:1708.07747.

Zhang, Z. and Sabuncu, M. (2018). Generalized cross entropy loss for training deep neural networks with noisy labels. Advances in neural information processing systems, 31.