Partial Label Learning Based on Fully Connected Deep Neural Network

被引：0
作者：
Li H. [1 ]
Wu L. [1 ]
He J. [1 ]
Zheng R. [1 ]
Zhou Y. [1 ]
Qiao S. [2 ]
机构：
[1] College of Information and Communication Engineering, Dalian Minzu University, Dalian
[2] School of Management, Dalian Polytechnic University, Dalian
来源：
International Journal of Circuits, Systems and Signal Processing | 2022年 / 16卷
基金：
中国国家自然科学基金;
关键词：
Deep neural network; Partial label learning; Weakly annotated data;
D O I：
10.46300/9106.2022.16.35
中图分类号：
学科分类号：
摘要：
The ambiguity of training samples in the partial label learning framework makes it difficult for us to develop learning algorithms and most of the existing algorithms are proposed based on the traditional shallow machine learning models, such as decision tree, support vector machine, and Gaussian process model. Deep neural networks have demonstrated excellent performance in many application fields, but currently it is rarely used for partial label learning frame-work. This study proposes a new partial label learning algorithm based on a fully connected deep neural network, in which the relationship between the candidate labels and the ground-truth label of each training sample is established by defining three new loss functions, and a regularization term is added to prevent overfitting. The experimental results on the controlled U-CI datasets and real-world partial label datasets reveal that the proposed algorithm can achieve higher classification accuracy than the state-of-the-art partial label learning algorithms. © 2022, North Atlantic University Union NAUN. All rights reserved.
引用
收藏
页码：287 / 297
页数：10
相关论文
共 54 条
  • [11] Yu F., Zhang M.-L., Decision tree ensem-ble based partial label learning algorithm, Pattern Recognition and Artificial Intelligence, 29, 4, pp. 367-375, (2016)
  • [12] Wang D.-B., Li L., Zhang M.-L., Adaptive graph guided disambiguation for partial label learn-ing, Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 83-91, (2019)
  • [13] Carrara F., Falchi F., Caldelli R., Amato G., Becarelli R., Adversarial image detection in deep neural networks, Multimedia Tools and Applications, 78, 3, pp. 2815-2835, (2019)
  • [14] Seki H., Yamamoto K., Akiba T., Nak-agawa S., Discriminative learning of filterbank layer within deep neural network based speech recognition for speaker adaptation, IEICE Transactions on Information and Systems, 102, 2, pp. 364-374, (2019)
  • [15] Harrouk A. I., Barbar A. M., A psycholinguistic approach to career selection using nlp with deep neural network classifiers, IEEE International Multi-disciplinary Conference on Engineering Technology, pp. 1-6, (2018)
  • [16] Tang X., Wang H., Jiao C., Multi-objective evolutionary metric learning for image retrieval using convolutional neural network features, 2019 IEEE Congress on Evolutionary Computation, pp. 578-585, (2019)
  • [17] Tang X., Liu C., Ma J., Zhang X., Jiao L., Large-scale remote sensing image retrieval based on semi-supervised adversarial hashing, Remote Sens-ing, 11, 17, (2019)
  • [18] Montufar G., Pascanu R., Cho K., Bengio Y., On the number of linear regions of deep neural net-works, Advances in Neural Information Processing Systems, pp. 2924-2932, (2014)
  • [19] Cohen N., Sharir O., Shashua A., On the ex-pressive power of deep learning: A tensor analysis, Conference on Learning Theory, pp. 698-728, (2016)
  • [20] Bentkamp A., Blanchette J., Klakow D., A formal proof of the expressiveness of deep learn-ing, Journal of Automated Reasoning, 63, 2, pp. 347-368, (2019)
123456