Abstract Training deep neural network models comes with a heavy labeling cost. To reduce the cost, a deep transfer active learning method combining source domain and target domain is proposed. With the initial model transferred from source task, the current task samples with larger contribution to the model performance improvement are labeled by using a dynamical weighting combination of source domain difference and target domain uncertainty. Information extraction ratio(IER) is concretely defined in the specific case. An IER-based batch training strategy and a T&N batch training strategy are proposed to deal with model training process. The proposed method is tested on two cross-dataset transfer learning experiments. The results show that the transfer active learning method achieves good performance and reduces the cost of annotation effectively and the proposed strategies optimize the distribution of computing resources during the active learning process. Thus, the model learns more times from samples in the early phases and less times in the later and end phases.
Fund:Science and Technology Foundation of Guizhou Province(No.QKHJC[2018]1114)
Corresponding Authors:
CAO Yongfeng, Ph.D., professor. His research interests include pattern recognition.
About author:: LIU Dapeng, master student. His research interests include image processing and machine vision. SU Caixia, master, lecturer. Her research interests include remote sensing image processing. ZHANG Lun, master student. His research interests include image processing and machine vision.
LIU Dapeng,CAO Yongfeng,SU Caixia等. Deep Transfer Active Learning Method Combining Source Domain Difference and Target Domain Uncertainty[J]. , 2021, 34(10): 898-908.
[1] LECUN Y, BENGIO Y, HINTON G. Deep Learning. Nature, 2015, 521(7553): 436-444. [2] WEI Q D, SHAO F J, LIU J. Research Summary of Convolution Neural Network in Image Recognition // Proc of the International Conference on Data Processing and Applications. New York, USA: ACM, 2018: 39-44. [3] CHEN H S, LIU X R, YIN D W, et al. A Survey on Dialogue Systems: Recent Advances and New Frontiers. ACM SIGKDD Explorations Newsletter, 2017, 19(2): 25-35. [4] TAN C Q, SUN F C, KONG T, et al. A Survey on Deep Transfer Learning // Proc of the International Conference on Artificial Neural Networks. Berlin, Germany: Springer, 2018: 270-279. [5] HUANG S J, ZHAO J W, LIU Z Y. Cost-Effective Training of Deep CNNs with Active Model Adaptation // Proc of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mi-ning. New York, USA: ACM, 2018: 1580-1588. [6] OQUAB M, BOTTOU L, LAPTEV I, et al. Learning and Trans-ferring Mid-level Image Representations Using Convolutional Neural Networks // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2014: 1717-1724. [7] DENG J, DONG W, SOCHER R, et al. ImageNet: A Large-Scale Hierarchical Image Database // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2009: 248-255. [8] TZENG E, HOFFMAN J, ZHANG N, et al. Deep Domain Confusion: Maximizing for Domain Invariance[C/OL]. [2021-01-03]. https://arxiv.org/pdf/1412.3474v1.pdf. [9] BORGWARDT K M, GRETTON A, RASCH M J, et al. Integrating Structured Biological Data by Kernel Maximum Mean Discrepancy. Bioinformatics, 2006, 22(14): e49-e57. [10] LONG M S, CAO Y, WANG J M, et al. Learning Transferable Features with Deep Adaptation Networks // Proc of the 32nd International Conference on Machine Learning. New York, USA: ACM, 2015: 97-105. [11] BEN-DAVID S, BLITZER J, CRAMMER K, et al. A Theory of Learning from Different Domains. Machine Learning, 2010, 79(1): 151-175. [12] TZENG E, HOFFMAN J, SAENKO K, et al. Adversarial Discri-minative Domain Adaptation // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2017: 2961-2971. [13] GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Ge-nerative Adversarial Nets // Proc of the 27th International Confe-rence on Neural Information Processing Systems. Cambridge, USA: The MIT Press, 2014, II: 2672-2680. [14] SETTLES B. Active Learning Literature Survey. TR 1648. Madison, USA: University of Wisconsin-Madison, 2010. [15] SMAILAGIC A, COSTA P, NOH H Y, et al. Medal: Accurate and Robust Deep Active Learning for Medical Image Analysis // Proc of the 17th IEEE International Conference on Machine Lear-ning and Applications. Washington, USA: IEEE, 2018: 481-488. [16] WANG K Z, ZHANG D Y, LI Y, et al. Cost-Effective Active Learning for Deep Image Classification. IEEE Transactions on Circuits and Systems for Video Technology, 2017, 27(12): 2591-2600. [17] ZHOU Z W, SHIN J, ZHANG L, et al. Fine-Tuning Convolutional Neural Networks for Biomedical Image Analysis: Actively and Incrementally // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2017: 4761-4772. [18] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet Cla-ssification with Deep Convolutional Neural Networks[C/OL]. [2021-01-03]. http://www.cs.toronto.edu/~hinton/absps/imagenet.pdf. [19] DENG C, XUE Y M, LIU X L, et al. Active Transfer Learning Network: A Unified Deep Joint Spectral-Spatial Feature Learning Model for Hyperspectral Image Classification. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(3): 1741-1754. [20] 杨 菊.主动学习停止准则与评价测度研究.硕士学位论文.镇江:江苏科技大学, 2016. (YANG J. Study of Stopping Criteria and Performance Evaluation Metrics in Active Learning. Master Dissertation. Zhenjiang, China: Jiangsu University of Science and Technology, 2016.)
2021, Vol. 34 
