Abstract
In multi-label classification applied to music auto-tagging, the classification threshold is simply set to a constant value (called static threshold), which is usually unsuitable for the classification on imbalanced datasets. There are many approaches to solve this problem. Some find an appropriate threshold for the whole dataset, while the others find one for each tag or for each individual musical instance. In this paper, we present a method for finding an appropriate classification threshold for each individual track using multiple techniques. The ranking model used to experiment with the thresholding model is built based on fully convolutional neural network structure. The performance of the classifier including the thresholding strategy is evaluated against the classifier using static threshold on various evaluation metrics. The results show that the proposed method helps to improve the classification quality of classifier to testing instances.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsReferences
Bertin-Mahieux, T., et al.: The million song dataset. In: Proceedings of the 12th International Conference on Music Information Retrieval (ISMIR 2011) (2011)
Chen, B., Gu, W., Hu, J.: An improved multi-label classification method and its application to functional genomics. Int. J. Comput. Biol. Drug Des. (IJCBDD) 3(2), 133–145 (2010)
Choi, K., et al.: Transfer learning for music classification and regression tasks. CoRR abs/1703.09179 (2017). arXiv:1703.09179
Defferrard, M., et al.: FMA: a dataset for music analysis. In: 18th International Society for Music Information Retrieval Conference (2017). https://arxiv.org/abs/1612.01840
Fan, R.-E., Lin, C.-J.: A Study on Threshold Selection for Multi-label Classification. Department of Computer Science, National Taiwan University (2005)
Law, E., et al.: Evaluation of algorithms using games: the case of music tagging. In: Proceedings of the 10th International Society for Music Information Retrieval Conference, ISMIR 2009, pp. 387–392 (2009). ISBN 9780981353708. English (US)
LeCun, Y., Haffner, P., Bottou, L., Bengio, Y.: Object recognition with gradient-based learning. Shape, Contour and Grouping in Computer Vision. LNCS, vol. 1681, pp. 319–345. Springer, Heidelberg (1999). https://doi.org/10.1007/3-540-46805-6_19
Al-Otaibi, R., Flach, P., Kull, M.: Multi-label classification: a comparative study on threshold selection methods (2014)
Zhang, M., Zhou, Z.: A review on multi-label learning algorithms. IEEE Trans. Knowl. Data Eng. 26(8), 1819–1837 (2014). https://doi.org/10.1109/TKDE.2013.39. ISSN: 1041-4347
Zou, Q.: Finding the best classification threshold in imbalanced classification. Big Data Res. 5, 2–8 (2016)
Acknowledgements
The authors would like to thank Faculty of Computer Science and Engineering, HCMC University of Technology for providing computing facilities to this study. The experiments presented in this paper are tested on the High Performance Computing Lab (HPC Lab) of the faculty.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer Nature Switzerland AG
About this paper
Cite this paper
Nguyen Cao Minh, K., Dang An, T., Tran Quang, V., Tran, V.H. (2018). Comparative Study on Different Approaches in Optimizing Threshold for Music Auto-Tagging. In: Dang, T., Küng, J., Wagner, R., Thoai, N., Takizawa, M. (eds) Future Data and Security Engineering. FDSE 2018. Lecture Notes in Computer Science(), vol 11251. Springer, Cham. https://doi.org/10.1007/978-3-030-03192-3_18
Download citation
DOI: https://doi.org/10.1007/978-3-030-03192-3_18
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-03191-6
Online ISBN: 978-3-030-03192-3
eBook Packages: Computer ScienceComputer Science (R0)