Orthogonality is Better: Auxiliary Problems in ASO Algorithm[J]. Chinese Journal of Electronics, 2012, 21(4): 645-650.
Citation: Orthogonality is Better: Auxiliary Problems in ASO Algorithm[J]. Chinese Journal of Electronics, 2012, 21(4): 645-650.

Orthogonality is Better: Auxiliary Problems in ASO Algorithm

  • Received Date: 2011-07-01
  • Rev Recd Date: 2012-02-01
  • Publish Date: 2012-10-25
  • We propose a principle called orthogonality for Auxiliary problems (APs) selection in Alternating structure optimization (ASO) algorithm. Both theoretical analyses and experimental results indicate the following conclusions. If the weight matrices of different types of APs are orthogonal or approximately orthogonal, their multi-combinations perform better than or equal to any components. Moreover, as long as the ratios of their components are appropriate, even if the total amounts of APs are fixed, the multi-combinations still perform better than or equal to any components. In short, the principle of orthogonality holds.
  • loading
  • Y. Zhong, “A unified theory of information, knowledge andintelligence”, Chinese Journal of Electronics, Vol.12, No.3,pp.391-396, 2003.
    Y. Zhong, “Consciousness machines: theory and applications”,Chinese Journal of Electronics, Vol.6, No.1, pp.42-45, 1997.
    R.K. Ando, T. Zhang, “A framework for learning predictivestructures from multiple tasks and unlabeled data”, Journal ofMachine Learning Research, Vol.6, No.11, pp.1817-1853, 2005.
    R.K. Ando, T. Zhang, “A high-performance semi-supervisedlearning method for text chunking”, Proc. of the 43rd AnnualMeeting of the Association for Computational Linguistics, AnnArbor, Michigan, USA, pp.1-9, 2005.
    J. Chen, L. Tang, J. Liu and J. Ye, “A convex formulation forlearning shared structures from multiple tasks”, Proc. of the26th International Conference on Machine Learning, Montreal,Canada, pp.137-144, 2009.
    X. Bai, T. Zhang, S. He and X.Wang, “Chinese syntactic chunkingbased on ASO algorithm”, Proc. of the 13th China NationalConference on Artificial Intelligence, Beijing, China, 2009.
    C. Liu, H.T. Ng, “Learning predictive structures for semanticrole labeling of NomBank”, Proc. of the 45th Annual Meetingof the Association for Computational Linguistics, Prague,Czech Republic, pp.208-215, 2007.
    S. He, T. Zhang, X. Wang, X. Bai and Y. Dong, “Incorporatingmulti-task learning in conditional random fields for chunkingin semantic role labeling”, Proc. of International Conferenceon Natural Language Processing and Knowledge Engineering,Dalian, China, pp.47-51, 2009.
    R.K. Ando, “Applying alternating structure optimization toword sense disambiguation”, Proc. of the 10th Conferenceon Computational Natural Language Learning, New York City,USA, pp.77-84, 2006.
    T. Zhang, X. Wang, X. Bai and S. He, “Relevancy of auxiliaryproblems in alternating structure optimization algorithm”,Journal of Beijing University of Posts and Telecommunications,Vol.34, No.4, pp.14-18, 2011.
    G.H. Golub, C.F.V. Loan, Matrix Computations (3rd edn),Johns Hopkins University Press, Baltimore, Maryland, USA,1996.
    T. Zhang, X. Wang and H. Tong, “Researches on combinationsof auxiliary problems in ASO (Alternating Structure Optimization)algorithm”, Proc. of the 2011 International Conferenceon Future Computer Science and Education, Xi’an, China,pp.608-614, 2011.
    T. Zhang, X. Wang and H. Tong, “Domain adaptation in Alternatingstructure optimization (ASO) algorithm”, Proc. ofthe 11th IASTED International Conference on Artificial Intelligenceand Applications, Innsbruck, Austria, pp.50-55, 2011.
    T. Zhang, X.Wang, H. Tong and Y. Zhong, “Auxiliary problemsselection in semantic role labeling”, Journal of ComputationalInformation Systems, Vol.8, No.2, pp.549-561, 2012.
  • 加载中

Catalog

    通讯作者: 陈斌, bchen63@163.com
    • 1. 

      沈阳化工大学材料科学与工程学院 沈阳 110142

    1. 本站搜索
    2. 百度学术搜索
    3. 万方数据库搜索
    4. CNKI搜索

    Article Metrics

    Article views (363) PDF downloads(999) Cited by()
    Proportional views
    Related

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return