(34.237.124.210) 您好!臺灣時間:2021/02/25 18:21
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:余卓涵
研究生(外文):Cho-Han Yu
論文名稱:一種基於錯誤校正之漸進式學習演算法
論文名稱(外文):An Incremental Learning Algorithm Based on Error Corrections
指導教授:李新林李新林引用關係
指導教授(外文):SingLing Lee
口試委員:何漢彰李新林吳昇江為國
口試委員(外文):Hann-Jang HoSingLing LeeSun WuWei-Kuo Chiang
口試日期:2012-03-19
學位類別:碩士
校院名稱:國立中正大學
系所名稱:資訊工程研究所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2012
畢業學年度:100
語文別:英文
論文頁數:31
中文關鍵詞:漸進式學習增量學習資料串流
外文關鍵詞:incremental learningdata stream
相關次數:
  • 被引用被引用:0
  • 點閱點閱:245
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:9
  • 收藏至我的研究室書目清單書目收藏:0
Incremental learning is characterized by the use of new training samples to adjust the current classifier without fully retraining. It is suitable for classification problems with frequent updates. Incremental SVM learning is a common algorithm of incremental learning, but it is not designed for imbalanced class distribution problems. Thus, when the class distribution is imbalanced, the learning algorithm needs larger amounts of training samples to establish a slightly accurate classifier. In this paper, an efficient and stable incremental learning algorithm is proposed. This learning algorithm is based on the concept of online passive aggressive algorithm but adds minimizing the error summation of the training samples into its original optimization problem. The experimental results conducted in this paper show that unstable problems caused by noisy datasets are apparently improved using the proposed algorithm in comparison with online passive aggressive algorithm. Moreover, in dealing with imbalanced datasets, the new algorithm which is able to achieve high classification accuracy during the early runs (run 1 to 3) outperforms incremental SVM learning which normally needs over five runs to see some effects.
1 Introduction .................................... 1
2 Related Works ................................... 5
2.1 Online Learning Algorithm ..................... 5
2.2 Perceptron Learning Algorithm ................. 6
2.3 Online Passive Aggressive Algorithm ........... 9
3 Incremental Passive-Aggressive Learning ........ 12
3.1 Update Model of Incremental PA Learning....... 12
3.2 Constrained Optimization ..................... 14
3.3 Classifier Selection ......................... 15
3.4 Adapted for Imbalanced Dataset ............... 17
4 Experiments .................................... 18
4.1 Experimental Environment...................... 18
4.1.1 Datasets ................................... 18
4.1.2 The Accuracy of Classification ............. 19
4.1.3 Experimental Design ........................ 19
4.1.4 Experimental Setting ....................... 21
4.2 Experimental Results ......................... 22
5 Conclusion ..................................... 29
[1] Mohamed Medhat Gaber, Arkady Zaslavsky, Shonali Krishnaswamy, “Mining Data Streams: A Review,” in ACM SIGMOD Record, Vol. 34, ISSN: 0163-5808, 2005.

[2] Rosenblatt, Frank, “The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain, Cornell Aeronautical Laboratory,” Psychological Review, v65, No. 6, pp. 386–408, 1958.

[3] Koby Crammer, Ofer Dekel, Joseph Keshet, Shai Shalev-Shwartz, Yoram Singer, “Online Passive-Aggressive Algorithms,” Journal of Machine Learning Research 7 , pp. 551–585, 2006.

[4] John Ross Quinlan, “Induction of decision trees,” Machine Learning, 1, pp. 81-106, 1986.

[5] Koichiro Yamauchi, Nobuhiko Yamaguchi, Naohiro Ishii, “Incremental learning methods with retrieving of interfered pattern”, IEEE Transactions on Neural Networks, Vol. 10, pp. 1351, 1999.

[6] Rong Xiao, Jicheng Wang, Fuyan Zhang, “An approach to incremental SVM learning algorithm,” Proc. Twelfth IEEE Internat. Conf. on Tools with Artificial Intelligence (ICTAI 2000), pp. 268–273. 2000.

[7] Gert Cauwenberghs, Tomaso Poggio, “Incremental and Decremental Support Vector Machine Learning,” Adv. Neural Information Processing Systems (NIPS*2000), Cambridge MA: MIT Press, vol. 13, 2001.

[8] Nick Littlestone. “Learning Quickly When Irrelevant Attributes Abound: A New Linear-threshold Algorithm,” Machine Learning pp. 285–318, 1988.

[9] Michael Collins, “Discriminative Training Methods for Hidden Markov Models: Theory and Experiments with Perceptron Algorithms,” Proceedings of EMNLP, 2002.

[10] Yoav Freund, Robert E. Schapire, “Large margin classification using the perceptron algorithm,” Machine Learning, 37(3), pp. 277–296, 1999.

[11] Nick Littlestone, Manfred K. Warmuth, “The Weighted Majority Algorithm,” Information and Computation, 108(2), pp. 212-261, 1994.

[12] Po-Ching Lin, Tsung-Hsing Yen, Jui-Hsi Fu, Cho-Han Yu, “Analyzing Anomalous Spamming Activities in a Campus Network,” Proceedings of TANET Conference, 2011.

[13] Arthur Asuncion, David J. Newman, UCI machine learning repository, 2007. URL http://archive.ics.uci.edu/ml/
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
系統版面圖檔 系統版面圖檔