請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/68227
完整後設資料紀錄
DC 欄位 | 值 | 語言 |
---|---|---|
dc.contributor.advisor | 廖世偉 | |
dc.contributor.author | Sipun Kumar Pradhan | en |
dc.contributor.author | 施庫馬 | zh_TW |
dc.date.accessioned | 2021-06-17T02:15:12Z | - |
dc.date.available | 2027-10-24 | |
dc.date.copyright | 2018-03-01 | |
dc.date.issued | 2017 | |
dc.date.submitted | 2017-10-23 | |
dc.identifier.citation | 1. Zaremba, W. and I. Sutskever, Learning to execute. preprint arXiv:1410.4615, 2014.
2. Wei, C., et al., Unbiased online active learning in data streams. In Proceedings of the seventeenth ACM sigkdd conference (KDD), January 2011. 3. Vlachos, A., A stopping criterion for active learning. Computer Speech and Language, 22(3):295–312, 2008. 4. Sutton, R. and A.G. Barto, Reinforcement Learning: An Introduction, MIT Press. 1998. 5. Seung, H.S., M. Opper, and H. Sompolinsky, Query by committee. In Proceedings of the ACM Workshop on Computational Learning Theory, pages 287–294, 1992. 6. Settles, B., M. Craven, and S. Ray, Multiple-instance active learning. In Advances in Neural Information Processing Systems (NIPS), 2008. volume 20: p. pages 1289– 1296. 7. Settles, B. and M. Craven, An analysis of active learning strategies for sequence labeling tasks. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), 2008(ACL Press): p. pages 1069–1078. 8. Sepp, H., Y.A. Steven, and C.P. R., Learning to learn using gradient descent. In Artificial Neural NetworksICANN, 2001: pp. 87–94. 9. Sepp, H. and S. Jürgen, Long short-term memory. Neural computation, November 1997. 9(8):1735–1780. 10. Scheffer, T., C. Decomain, and S. Wrobel, Active hidden Markov models for information extraction. In Proceedings of the International Conference on Advances in Intelligent Data Analysis (CAIDA), 2001(Springer-Verlag): pages 309–318. 11. P., K.D., et al., Semisupervised learning with deep generative models. In Proc. of the conference on neural information processing systems, NIPS, 2014. 12. Oriol, V., et al., Matching networks for one shot learning. In Proc. of the conference on neural information processing systems, NIPS, 2016. 13. Mark, W. and F. Chelsea, Active One-shot Learning. NIPS, 2016. 14. Lewis, D. and W. Gale, A sequential algorithm for training text classifiers. In Proceedings of the ACM SIGIR Conference on Research and Development in Information Retrieval, 1994. ACM/Springer: p. 3–12. 15. Lewis, D. and J. Catlett, Heterogeneous uncertainty sampling for supervised learning. In Proceedings of the International Conference on Machine Learning (ICML), 1994: p. 148–156. 16. Lake, B.M., R. Salakhutdinov, and J.B. Tenenbaum, Human-level concept learning through probabilistic program induction. Science, 2015. 350(6266): p. 1332-1338. 17. Krogh, A., V. Jesper, and e. al, Neural network ensembles, cross validation, and active learning. Advances in neural information processing systems, 1995. 7:231–238. 18. Goodman, J., Exponential priors for maximum entropy models. In Proceedings of Human Language Technology and the North American Association for Computational Linguistics (HLT-NAACL), 2004. ACL Press: p. pages 305–312. 19. Culotta, A. and A. McCallum, Reducing labeling effort for stuctured prediction tasks. In Proceedings of the National Conference on Artificial Intelligence (AAAI), 2005. AAAI Press: p. Pages:746–751. 20. Chen, S.F. and R. Rosenfeld, A survey of smoothing techniques for ME models. IEEE Transactions on Speech and Audio Processing, 2000. 8(1):37–50. 21. Burr, S., Active Learning Literature Survey. Computer Sciences Technical Report 1648, 2009. 22. Bloodgood, M. and V. Shanker, A method for stopping active learning based on stabilizing predictions and the need for user-adjustable stopping. In Proceedings of the Conference on Natural Language Learning (CoNLL), 2009. ACL Press: p. pages 39–47. 23. Antti, R., et al., Semisupervised learning with ladder network. In Proc. of the conference on neural information processing systems, NIPS, 2015. 24. Alex, G., W. Greg, and D. Ivo, Neural turing machines. arXiv preprint arXiv:1410.5401, 2014. 25. Adam, S., et al., One-shot learning with memory-augmented neural networks. ICML, 2016. 26. Volodymyr, M., et al., Playing atari with deep reinforcement learning. In Proc. of the conference on neural information processing systems, NIPS, 2013. 27. Mikolov, T., et al., Recurrent neural network based language model. in Interspeech, 2010: p. p. 3. 28. Graves, A., Supervised Sequence Labelling with Recurrent Neural Networks, Technical University of Munich, 2008. | |
dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/68227 | - |
dc.description.abstract | Interactive selection of desired training samples for labeling from a hay stack of unlabeled examples is an extremely challenging task in supervised learning. In many real world applications, active selection of training examples can significantly reduce the number of labelled training examples to learn a classification function. Different strategies of active learning have been proposed that iteratively select a single new example from a set of unlabeled examples, query the corresponding class label and then perform retraining of the current classifier. However, to reduce computational time for training, it might be necessary to select batches of new training examples instead of single examples. My research goal in this thesis is to develop learning models that can automatically learn new facts to optimize selection learning without having to be re-trained will full corpus. Our method applies the active learning methodology, and the user only needs to label minimal initial training data and subsequent query data.
I investigate a new class of learning models called active learning with few shot learning. The main advantage of this framework is that it requires little feature engineering and domain specificity whilst matching or surpassing state-of-the-art results. Furthermore, it can easily be trained to be used with any kind of Open-domain. The premise of active learning is that there are costs associated with labeling and with making an incorrect prediction. Reinforcement learning allows for the explicit specification of those costs, and directly finds a labelling policy to optimize those costs. Finally, I show that with memory augmentation our model can reach promising and learn to perform non-trivial operations. I confirm those results by comparing my system to various well-crafted baseline Datasets and future work is discussed. | en |
dc.description.provenance | Made available in DSpace on 2021-06-17T02:15:12Z (GMT). No. of bitstreams: 1 ntu-106-R04922153-1.pdf: 1248505 bytes, checksum: 8dd4bbd826288938dce9737401413d75 (MD5) Previous issue date: 2017 | en |
dc.description.tableofcontents | Contents
Acknowledgement…………………………………………………………………………...i Abstract……………………………………………………………………………………...ii List of Figures……………………………………………………………………………….v List ofTables………………………………………………………………………………..vi Chapter 1 Introduction 1 1.1 Overview 1 Learning from sparse data: 1 Long-term dependency: 1 1.2 Motivation 2 1.3 Brief Literature 3 1.4 Contributions 5 1.5 Thesis Organization 6 Chapter 2 Active learning Background 7 2.1 Active learning and Artificial Intelligence 7 2.2 Reinforcement Learning Model 8 2.3 Why Deep Reinforcement learning? 10 2.4 Learning Multiple Levels of Inference 10 2.5 Long Short Term Memory 11 2.5.1 The LSTM Architecture 12 2.6 Meta-Learning Task Methodology 15 Chapter 3 Active Learning Strategy Frameworks 17 3.1 Uncertainty Sampling 17 3.2 Query-By-Committee 18 3.3 Expected Model Change 18 3.4 Stopping Criteria 19 Chapter 4 Implementation 21 4.1 Task Methodology 21 4.2 Reinforcement Learning Model 22 4.3 Reading and Writing to the memory 23 4.4 Least Recently Used Access 25 Chapter 5 Experimental result 28 5.1 Dataset 28 5.2 Preprocessing 29 5.3 Results 29 5.3.1 Human baseline 29 Chapter 6 Conclusion 33 6.1 Main Contributions 33 6.2 Future work 34 Reference………………………………………………………………………………...…35 | |
dc.language.iso | en | |
dc.title | 記憶增強的主動少量學習 | zh_TW |
dc.title | Memory-Augmented Active Few-shot Learning | en |
dc.type | Thesis | |
dc.date.schoolyear | 106-1 | |
dc.description.degree | 碩士 | |
dc.contributor.oralexamcommittee | 周碩彥,黃思皓,葉羅堯,蘇中才 | |
dc.subject.keyword | Active learning,One-shot Learning,Memory-Augmented Neural Networks., | zh_TW |
dc.relation.page | 38 | |
dc.identifier.doi | 10.6342/NTU201704312 | |
dc.rights.note | 有償授權 | |
dc.date.accepted | 2017-10-24 | |
dc.contributor.author-college | 電機資訊學院 | zh_TW |
dc.contributor.author-dept | 資訊工程學研究所 | zh_TW |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-106-1.pdf 目前未授權公開取用 | 1.22 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。