Label Distribution Learning Based on Ensemble Neural Networks
Label distribution learning (LDL), as an extension of multi-label learning, is a new arising machine learning technique to deal with label ambiguity problems. The maximum entropy model is commonly used in label distribution learning. However, it does not consider the correlation between the labels and is not suitable for nonlinear relationships, and the prediction performance is also limited. In this paper, we propose a label distribution learning algorithm based on ensemble neural networks. The algorithm trains neural networks with preferences using training sets with different label sets to construct base learners, and combines the base learners with the weights, which is learned by the combined learner to obtain the final learning results. Experimental results show that the proposed algorithm is effective for label distribution data.
KeywordsLabel distribution learning Neural networks Ensemble learning Maximum entropy model
This work was partially supported by the National Natural Science Foundation of China (Nos. 61473259, 61502335, 61070074, 60703038) and the Hunan Provincial Science & Technology Program Project (No. 2018TP1018).
- 1.Berger, A.L., Pietra, S.A.D., Pietra, V.J.D.: A maximum entropy approach to natural language processing. Comput. Linguist. 22, 39–71 (1996)Google Scholar
- 4.Geng, X., Xia, Y.: Head pose estimation based on multivariate label distribution. In: Computer Vision and Pattern Recognition, pp. 1837–1842 (2014)Google Scholar
- 5.Geng, X., Xu, N., Sao, R.: Label enhancement for label distribution learning. J. Comput. Res. Dev. 54(6), 1171–1184 (2017)Google Scholar
- 7.Glorot, X., Bordes, A., Bengio, Y.: Deep sparse rectifier neural networks. In: International Conference on Artificial Intelligence and Statistics, pp. 315–323 (2012)Google Scholar
- 9.Kingma, D.P., Ba, J.: Adam: a method for stochastic optimization. CoRR abs/1412.6980 (2014). http://arxiv.org/abs/1412.6980
- 10.Sao, D., Yang, W.Y., Zhao, H.: Application of k-means algorithm to realize label distribution learning. Trans. Intell. Syst. 12(3), 325–332 (2017)Google Scholar
- 11.Sung-Hyuk, C.: Comprehensive survey on distance/similarity measures between probability density functions. Int. J. Math. Model. Methods Appl. Sci. 1(4), 300–307 (2007)Google Scholar
- 12.Wang, Q., Geng, X.: Face estimation based on adaptive label distribution. Master’s thesis, Southeast University, NanJing (2015)Google Scholar
- 16.Zhao, Q., Geng, X.: Study on several problems of label distribution learning. Master’s thesis, Southeast University, NanJing (2016)Google Scholar
- 17.Zhao, Q., Geng, X.: Selection of objective function in label distribution learning. J. Front. Comput. Sci. Technol. 11(5), 708–719 (2017)Google Scholar
- 18.Zhou, Y., Geng, X.: Application of label distribution in face image feature recognition. Master’s thesis, Southeast University, NanJing (2016)Google Scholar
- 19.Zhou, Y., Xue, H., Geng, X.: Emotion distribution recognition from facial expressions. In: Proceedings of the 23rd ACM International Conference on Multimedia, pp. 1247–1250. ACM (2015)Google Scholar