Peng Liu
Skip Abstract Section
Abstract
Which samples should be labelled in a large dataset is one of the most important problems for the training of deep learning. So far, a variety of active sample selection strategies related to deep learning have been proposed in the literature. We defined them as Active Deep Learning (ADL) only if their predictor or selector is a deep model, where the basic learner is called the predictor and the labeling schemes are called the selector. In this survey, we categorize ADL into model-driven ADL and data-driven ADL by whether its selector is model driven or data driven. We also introduce the different characteristics of the two major types of ADL, respectively. We summarized three fundamental factors in the designation of a selector. We pointed out that, with the development of deep learning, the selector in ADL also is experiencing the stage from model driven to data driven. The advantages and disadvantages between data-driven ADL and model-driven ADL are thoroughly analyzed. Furthermore, different sub-classes of data-drive or model-driven ADL are also summarized and discussed emphatically. Finally, we survey the trend of ADL from model driven to data driven.
- [1] . 2016. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR’16). IEEE Computer Society, 770–778.Google Scholar
Cross Ref
- [2] . 2009. ImageNet: A large-scale hierarchical image database. In Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’09). IEEE Computer Society, 248–255.Google ScholarCross Ref
- [3] . 2017. SVM or deep learning? A comparative study on remote sensing image classification. Soft Comput. 21, 23 (2017), 7053–7065.Google ScholarDigital Library
- [4] . 2018. Remote sensing big data: Theory, methods and applications. Remote Sensing 10, 5 (2018), 711.Google Scholar
- [5] . 1994. Improving generalization with active learning. Mach. Learn. 15, 2 (1994), 201–221.Google ScholarCross Ref
- [6] . 2009. Active Learning Literature Survey.
Technical Report . University of Wisconsin—Madison Department of Computer Sciences.Google Scholar - [7] . 2010. Optimal experimental design. Wiley Interdisc. Rev.: Comput. Stat. 2, 5 (2010), 581–589.Google ScholarDigital Library
- [8] . 2009. A literature survey of active machine learning in the context of natural language processing. https://www.ccs.neu.edu/home/vip/teach/MLcourse/4_boosting/materials/SICS-T--2009-06--SE.pdf.Google Scholar
- [9] . 2016. A survey of active learning in collaborative filtering recommender systems. Comput. Sci. Rev. 20 (2016), 29–50.Google ScholarDigital Library
- [10] . 2011. A survey of active learning algorithms for supervised remote sensing image classification. IEEE J. Select. Top. Sign. Process. 5, 3 (2011), 606–617.Google ScholarCross Ref
- [11] . 2020. A survey of deep active learning. arXiv:2009.00236. Retrieved from https://arxiv.org/abs/2009.00236.Google Scholar
- [12] . 2020. A survey of active learning for text classification using deep neural networks.arXiv:2008.07267. Retrieved from https://arxiv.org/abs/2008.07267.Google Scholar
- [13] . 2021. A survey on active learning and human-in-the-loop deep learning for medical image analysis. Medical Image Analysis 71 (2021), 102062.Google Scholar
- [14] . 2018. A benchmark and comparison of active learning for logistic regression. Pattern Recogn. 83 (2018), 401–415. https://www.sciencedirect.com/science/article/pii/S0031320318302140.Google ScholarDigital Library
- [15] . 2019. Intelligent labeling based on fisher information for medical image segmentation using deep learning. IEEE Trans. Med. Imag. 38, 11 (2019), 2642–2653.Google ScholarCross Ref
- [16] . 2019. On the difficulty of warm-starting neural network training.
arxiv:1910.08475 . Retrieved from http://arxiv.org/abs/1910.08475.Google Scholar - [17] . 2019. Multi-criteria active deep learning for image classification. Knowl. Bas. Syst. 172 (2019), 86–94. Google ScholarDigital Library
- [18] . 2017. Active deep learning for classification of hyperspectral images. IEEE J. Select. Top. Appl. Earth Observ. Remote Sens. 10, 2 (
February 2017), 712–724.Google ScholarCross Ref - [19] . 2020. Deep active learning for surface defect detection. Sensors 20, 6 (2020).Google ScholarCross Ref
- [20] . 2017. Deep bayesian active learning with image data. arXiv:1703.02910. Retrieved from https://arxiv.org/abs/1703.02910.Google Scholar
- [21] . 2019. Learning loss for active learning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR’19). 93–102.Google ScholarCross Ref
- [22] . 2017. Active one-shot learning. arXiv:1702.06559. Retrieved from http://arxiv.org/abs/1702.06559.Google Scholar
- [23] . 2018. Meta-learning transferable active learning policies by deep reinforcement learning. arxXiv:1806.04798. Retrieved from http://arxiv.org/abs/1806.04798.Google Scholar
- [24] . 2018. Meta-learning for batch mode active learning. In Proceedings of the 6th International Conference on Learning Representations (ICLR’18). OpenReview.net.Google Scholar
- [25] . 2017. A meta-learning approach to one-step active learning. arXiv:1706.08334. Retrieved from http://arxiv.org/abs/1706.08334.Google Scholar
- [26] . 2017. Generative adversarial active learning. arXiv:1702.07956. Retrieved from http://arxiv.org/abs/1702.07956.Google Scholar
- [27] . 2020. Adversarial sampling for active learning. In Proceedings of the IEEE Winter Conference on Applications of Computer Vision. 3071–3079.Google ScholarCross Ref
- [28] . 2018. Adversarial active learning for deep networks: A margin based approach. arXiv:1802.09841. Retrieved from https://arxiv.org/abs/1802.09841.Google Scholar
- [29] . 2009. Multi-class active learning for image classification. In Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’09). IEEE Computer Society, 2372–2379. Google ScholarCross Ref
- [30] . 2020. Deep active learning in remote sensing for data efficient change detection. CoRR abs/2008.11201.Google Scholar
- [31] . 2017. Active learning for human pose estimation. In Proceedings of the IEEE International Conference on Computer Vision (ICCV’17). IEEE Computer Society, 4373–4382.Google ScholarCross Ref
- [32] . 2018. Query-by-committee improvement with diversity and density in batch active learning. Inf. Sci. 454-455 (2018), 401–418.Google ScholarCross Ref
- [33] . 2020. Hyperspectral image classification with convolutional neural network and active learning. IEEE Trans. Geosci. Remote Sens. (2020), 1–13. Google ScholarCross Ref
- [34] . 2018. MIDAS: Model-independent training data selection under cost constraints. IEEE Access 6 (2018), 74462–74474.Google ScholarCross Ref
- [35] . 2021. A novel framework for detecting social bots with deep neural networks and active learning. Knowl.-Bas. Syst. 211 (2021), 106525.Google ScholarCross Ref
- [36] . 2020. Cost sensitive active learning using bidirectional gated recurrent neural networks for imbalanced fault diagnosis. Neurocomputing 407 (2020), 232–245.Google ScholarCross Ref
- [37] . 2019. Active transfer learning network: A unified deep joint spectral-spatial feature learning model for hyperspectral image classification. IEEE Trans. Geosci. Remote. Sens. 57, 3 (2019), 1741–1754. Google ScholarCross Ref
- [38] . 2016. Stacked sparse autoencoder (SSAE) for nuclei detection on breast cancer histopathology images. IEEE Trans. Med. Imag. 35, 1 (2016), 119–130.Google ScholarCross Ref
- [39] . 2020. Joint temporal context exploitation and active learning for video segmentation. Pattern Recogn. 100 (2020), 107158.Google ScholarDigital Library
- [40] . 2019. Data stream classification using active learned neural networks. Neurocomputing 353 (2019), 74–82.
Recent Advancements in Hybrid Artificial Intelligence Systems. Google ScholarDigital Library - [41] . 2016. DeepFool: A simple and accurate method to fool deep neural networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR’16). IEEE Computer Society, 2574–2582. Google ScholarCross Ref
- [42] . 2017. A geometric approach to active learning for convolutional neural networks. arXiv: abs/1708.00489. Retrieved from https://arxiv.org/abs/1708.00489.Google Scholar
- [43] . 2017. Active learning for convolutional neural networks: A core-set approach. arXiv:1708.00489. Retrieved from https://arxiv.org/abs/1708.00489.Google Scholar
- [44] . 2020. Towards robust and reproducible active learning using neural networks (unpublished).Google Scholar
- [45] . 1992. On the Mathematical Foundations of Theoretical Statistics. Springer, New York, NY, 11–44.Google Scholar
- [46] . 2000. Statistical active learning in multilayer perceptrons. IEEE Trans. Neural Netw. Learn. Syst. 11, 1 (2000), 17–26.Google ScholarDigital Library
- [47] . 2000. The value of unlabeled data for classification problems. In Proceedings of the 17th International Conference on Machine Learning. Morgan Kaufmann, 1191–1198.Google Scholar
- [48] . 2008. Multiple-instance active learning. In Advances in Neural Information Processing Systems. MIT Press, 1289–1296.Google ScholarDigital Library
- [49] . 2009. Batch mode active learning with applications to text categorization and image retrieval. IEEE Trans. Knowl. Data Eng. 21, 9 (2009), 1233–1248.Google ScholarDigital Library
- [50] . 2015. Convergence rates of active learning for maximum likelihood estimation. In Advances in Neural Information Processing Systems 28: Annual Conference on Neural Information Processing Systems, , , and . (Eds.). 1090–1098.Google Scholar
- [51] . 2017. Asymptotic analysis of objectives based on fisher information in active learning. J. Mach. Learn. Res. 18 (2017), 34:1–34:41.Google Scholar
- [52] . 2017. Active discriminative text representation learning. In Proceedings of the 31st AAAI Conference on Artificial Intelligence, and (Eds.). AAAI Press, 3386–3392.Google ScholarCross Ref
- [53] . 2019. Deep batch active learning by diverse, uncertain gradient lower bounds. arXiv:1906.03671. Retrieved from http://arxiv.org/abs/1906.03671.Google Scholar
- [54] . 2015. Active learning for semi-supervised clustering based on locally linear propagation reconstruction. Neural Netw. 63 (2015), 170–184.Google ScholarDigital Library
- [55] . 2011. Active learning based on locally linear reconstruction. IEEE Trans. Pattern Anal. Mach. Intell. 33, 10 (2011), 2026–2038.Google ScholarDigital Library
- [56] . 2019. Rethinking deep active learning: Using unlabeled data at model training. arXiv:cs.CV/1911.08177. Retrieved from https://arxiv.org/abs/1911.08177.Google Scholar
- [57] . 1998. Employing EM and pool-based active learning for text classification. In Proceedings of the 15th International Conference on Machine Learning (ICML’98). Morgan Kaufmann, San Francisco, CA, 350–358.Google Scholar
- [58] . 2019. Single shot active learning using pseudo annotators. Pattern Recogn. 89 (2019), 22–31.Google ScholarCross Ref
- [59] . 2017. Suggestive annotation: A deep active learning framework for biomedical image segmentation. In Proceedings of the International Conference on Medical Image Computing and Computer-assisted Intervention. Springer, 399–407.Google ScholarDigital Library
- [60] . 2015. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR’15). IEEE Computer Society, 3431–3440.Google ScholarCross Ref
- [61] . 2017. Learning to diversify deep belief networks for hyperspectral image classification. IEEE Trans. Geosci. Remote. Sens. 55, 6 (2017), 3516–3530.Google ScholarCross Ref
- [62] . 2020. A divide-and-conquer approach to geometric sampling for active learning. Expert Syst. Appl. 140 (2020).Google ScholarDigital Library
- [63] . 2014. Bayesian active remote sensing image classification. IEEE Trans. Geosci. Remote Sens. 52, 4 (2014), 2186–2196.Google ScholarCross Ref
- [64] . 2015. An MRF model-based active learning framework for the spectral-spatial classification of hyperspectral imagery. IEEE J. Select. Top. Sign. Process. 9, 6 (2015), 1074–1088.Google ScholarCross Ref
- [65] . 2018. Active learning with convolutional neural networks for hyperspectral image classification using a new bayesian approach. IEEE Trans. Geosci. Remote Sens. 56, 11 (
Nov 2018), 6440–6461.Google ScholarCross Ref - [66] . 2015. Weight uncertainty in neural networks. arXiv:1505.05424. Retrieved from https://arxiv.org/abs/1505.05424.Google Scholar
- [67] . 2019. Bayesian batch active learning as sparse subset approximation. In Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems (NeurIPS’19), , , , , , and (Eds.). 6356–6367.Google Scholar
- [68] . 2020. Combining deep generative and discriminative models for bayesian semi-supervised learning. Pattern Recogn. 100 (2020), 107156.Google ScholarDigital Library
- [69] . 2020. Active learning for segmentation based on bayesian sample queries. Knowl.-Bas. Syst. (2020), 106531.Google Scholar
- [70] . 2019. Hyperspectral image classification using spectral-spatial features with informative samples. IEEE Access 7 (2019), 20869–20878.Google ScholarCross Ref
- [71] . 2020. Consistency-based semi-supervised active learning: Towards minimizing labeling cost. arXiv:cs.LG/1910.07153. Retrieved from https://arxiv.org/abs/1910.07153.Google Scholar
- [72] . 2020. Incremental learning model inspired in rehearsal for deep convolutional networks. Knowl.-Bas. Syst. 208 (2020), 106460.Google ScholarCross Ref
- [73] . 2019. Combining clustering and active learning for the detection and learning of new image classes. Neurocomputing 358 (2019), 150–165.Google ScholarDigital Library
- [74] . 2020. Multi-criteria online frame-subset selection for autonomous vehicle videos. Pattern Recogn. Lett. 133 (2020), 349–355.Google ScholarCross Ref
- [75] . 2020. Self-paced active learning for deep CNNs via effective loss function. Neurocomputing (2020).Google Scholar
- [76] . 2019. Inductive conformal predictor for convolutional neural networks: Applications to active learning for image classification. Pattern Recogn. 90 (2019), 172–182.Google ScholarDigital Library
- [77] . 2018. The power of ensembles for active learning in image classification. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 9368–9377.Google ScholarCross Ref
- [78] . 2017. The Mnist Database of Handwritten Digits.
Technical Report . Retrieved from http://yann.lecun.com/exdb/mnist.Google Scholar - [79] . 2009. Learning Multiple Layers of Features from Tiny Images.
Technical Report .Google Scholar - [80] . 2020. Deep active learning: Unified and principled method for query and training. In Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics (AISTATS’20), and (Eds.), Vol. 108. PMLR, 1308–1318.Google Scholar
- [81] . 2020. Meta-learning in neural networks: A survey. CoRR abs/2004.05439.Google Scholar
- [82] . 2020. Active one-shot learning by a deep q-network strategy. Neurocomputing 383 (2020), 324–335.Google ScholarDigital Library
- [83] . 2017. Learning algorithms for active learning. In Proceedings of the 34th International Conference on Machine Learning (ICML’17), and (Eds.), Vol. 70. PMLR, 301–310.Google Scholar
- [84] . 2016. Matching networks for one shot learning. In Advances in Neural Information Processing Systems 29: Annual Conference on Neural Information Processing Systems, , , , , and (Eds.). 3630–3638.Google Scholar
- [85] . 2015. Learning to compare image patches via convolutional neural networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR’15). 4353–4361.Google ScholarCross Ref
- [86] . 2015. Siamese neural networks for one-shot image recognition.Google Scholar
- [87] . 2018. Learning to compare: Relation network for few-shot learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 1199–1208.Google ScholarCross Ref
- [88] . 2017. Optimization as a model for few-shot learning. In Proceedings of the 5th International Conference on Learning Representations (ICLR’17). OpenReview.net.Google Scholar
- [89] . 2021. Air quality predictions with a semi-supervised bidirectional LSTM neural network. Atmos. Poll. Res. 12, 1 (2021), 328–339.Google ScholarCross Ref
- [90] . 2017. Learning active learning from data. In Advances in Neural Information Processing Systems 30: Annual Conference on Neural Information Processing Systems, and (Eds.). 4225–4235.Google Scholar
- [91] . 2016. Meta-learning with memory-augmented neural networks. In Proceedings of the 33nd International Conference on Machine Learning (ICML’16), and (Eds.), Vol. 48. 1842–1850.Google Scholar
- [92] . 2019. Augmented memory networks for streaming-based active one-shot learning. arXiv:1909.01757. Retrieved from http://arxiv.org/abs/1909.01757.Google Scholar
- [93] . 1984. Temporal credit assignment in reinforcement learning.Google Scholar
- [94] . 2017. Learning how to active learn: A deep reinforcement learning approach. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP’17), , , and (Eds.). Association for Computational Linguistics, 595–605. Google ScholarCross Ref
- [95] . 2015. Human-level control through deep reinforcement learning. Nature 518, 7540 (2015), 529–533.Google ScholarCross Ref
- [96] . 2016. Asynchronous methods for deep reinforcement learning. In Proceedings of the International Conference on Machine Learning. PMLR, 1928–1937.Google ScholarDigital Library
- [97] . 2013. Playing atari with deep reinforcement learning. arXiv:1312.5602. Retrieved from https://arxiv.org/abs/1312.5602.Google Scholar
- [98] . 2016. Dueling network architectures for deep reinforcement learning. In Proceedings of the International Conference on Machine Learning. PMLR, 1995–2003.Google ScholarDigital Library
- [99] . 2019. A reinforcement one-shot active learning approach for aircraft type recognition. IEEE Access 7 (2019), 147204–147214.Google ScholarCross Ref
- [100] . 2017. Diet networks: Thin parameters for fat genomics. In Proceedings of the 5th International Conference on Learning Representations (ICLR’17). OpenReview.net.Google Scholar
- [101] . 2019. RadGrad: Active learning with loss gradients.arXiv:1906.07838. Retrieved from http://arxiv.org/abs/1906.07838.Google Scholar
- [102] . 2017. Learning how to active learn: A deep reinforcement learning approach. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP’17), , , and (Eds.). Association for Computational Linguistics, 595–605.Google ScholarCross Ref
- [103] . 2020. On the sample complexity of the linear quadratic regulator. Found. Comput. Math. 20, 4 (2020), 633–679.Google ScholarDigital Library
- [104] . 2017. Imagination-augmented agents for deep reinforcement learning. In Annual Conference on Neural Information Processing Systems (NIPS’17), , , and (Eds.). 5690–5701.Google Scholar
- [105] . 2018. Sample-efficient reinforcement learning with stochastic ensemble value expansion. In Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems (NeurIPS’18), , , , , , and (Eds.). 8234–8244.Google Scholar
- [106] . 2018. Model-based value estimation for efficient model-free reinforcement learning. arXiv:0803.00101. Retrieved from http://arxiv.org/abs/1803.00101.Google Scholar
- [107] . 2016. Continuous deep q-learning with model-based acceleration. In Proceedings of the 33nd International Conference on Machine Learning (ICML’16), and (Eds.), Vol. 48. JMLR.org, 2829–2838.Google Scholar
- [108] . 2018. Model-ensemble trust-region policy optimization. In Proceedings of the 6th International Conference on Learning Representations (ICLR’18). OpenReview.net.Google Scholar
- [109] . 2018. Efficient active learning for image classification and segmentation using a sample selection and conditional generative adversarial network. In Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 580–588.Google ScholarCross Ref
- [110] . 2019. Active semi-supervised learning based on self-expressive correlation with generative adversarial networks. Neurocomputing 345 (2019), 103–113.Google ScholarDigital Library
- [111] . 2017. Active decision boundary annotation with deep generative models. In Proceedings of the IEEE International Conference on Computer Vision (ICCV’17). IEEE Computer Society, 5296–5305.Google ScholarCross Ref
- [112] . 2019. Bayesian generative active deep learning. In Proceedings of the 36th International Conference on Machine Learning (ICML’19)
, and (Eds.), Vol. 97. PMLR, 6295–6304.Google Scholar - [113] . 2019. An active learning approach for reducing annotation cost in skin lesion analysis. In Proceedings of the 10th International Workshop on Machine Learning in Medical Imaging (MLMI’19), and (Eds.), Vol. 11861. Springer, 628–636.Google ScholarDigital Library
- [114] . 2019. Skin lesion analysis toward melanoma detection 2018: A challenge hosted by the international skin imaging collaboration (ISIC). CoRR abs/1902.03368.Google Scholar
- [115] . 2019. Adversarial representation active learning. arXiv:cs.CV/1912.09720. Retrieved from https://arxiv.org/abs/1912.09720.Google Scholar
- [116] . 2019. Discriminative active learning. arXiv:1907.06347. Retrieved from https://arxiv.org/abs/1907.06347.Google Scholar
- [117] . 2019. Variational adversarial active learning. arXiv:1904.00370. Retrieved from http://arxiv.org/abs/1904.00370.Google Scholar
- [118] . 2020. Task-aware variational adversarial active learning.arxiv:2002.04709. Retrieved from https://arxiv.org/abs/2002.04709.Google Scholar
- [119] . 2016. DeepFool: A simple and accurate method to fool deep neural networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR’16). IEEE Computer Society, 2574–2582.Google ScholarCross Ref
- [120] . 2018. Ranking CGANs: Subjective control over semantic image attributes. In Proceedings of the British Machine Vision Conference BMVC’18)). BMVA Press, 131.Google Scholar
- [121] . 2021. Active deep learning for hyperspectral image classification with uncertainty learning. IEEE Geosci. Remote Sens. Lett. (2021).Google Scholar
- [122] 2015. The multimodal brain tumor image segmentation benchmark (BRATS). IEEE Trans. Med. Imag. 34, 10 (2015), 1993–2024.Google ScholarCross Ref
- [123] . 2010. A survey on transfer learning. IEEE Trans. Knowl. Data Eng. 22, 10 (2010), 1345–1359.Google ScholarDigital Library
- [124] . 2016. A survey of transfer learning. J. Big Data 3 (2016), 9.Google ScholarCross Ref
- [125] . 2020. Meta-learning in neural networks: A survey. CoRR abs/2004.05439.Google Scholar
- [126] . 2020. A comprehensive survey of neural architecture search: Challenges and solutions. CoRR abs/2006.02903.Google Scholar
- [127] . 2019. Deep metric learning: A survey. Symmetry 11, 9 (2019), 1066.Google ScholarCross Ref
- [128] . 2020. Generalizing from a few examples: A survey on few-shot learning. ACM Comput. Surv. 53, 3 (2020), 63:1–63:34.Google Scholar
- [129] . 2018. A brief introduction to weakly supervised learning. Natl. Sci. Rev. 5, 1 (2018), 44–53.Google ScholarCross Ref
Index Terms
- A Survey on Active Deep Learning: From Model Driven to Data Driven
Recommendations
An Online Business Process Model-driven Generator of the Conceptual Database Model
WIMS '18: Proceedings of the 8th International Conference on Web Intelligence, Mining and SemanticsThe paper presents an online two-phase business process model-driven generator of the conceptual database model. The generator is implemented as a web-based, platform-independent tool, in contrast to the existing tools that are dependent on some ...
Model Driven Validation of System Architectures
HASE '11: Proceedings of the 2011 IEEE 13th International Symposium on High-Assurance Systems EngineeringThe architecture is the basic structure of every system. The system architect is responsible for ensuring that it fits to the system requirements even if these requirements change according to new conditions during development process. Our approach ...
Model-driven engineering
During the last decade a new trend of approaches has emerged, which considers models not just documentation artefacts, but also central artefacts in the software engineering field, allowing the creation or automatic execution of software systems ...
Comments