Multiscale Bagging and Its Applications

この論文にアクセスする

この論文をさがす

著者

抄録

We propose multiscale bagging as a modification of the bagging procedure. In ordinary bagging, the bootstrap resampling is used for generating bootstrap samples. We replace it with the multiscale bootstrap algorithm. In multiscale bagging, the sample size <i>m</i> of bootstrap samples may be altered from the sample size <i>n</i> of learning dataset. For assessing the output of a classifier, we compute bootstrap probability of class label; the frequency of observing a specified class label in the outputs of classifiers learned from bootstrap samples. A scaling-law of bootstrap probability with respect to σ<sup>2</sup>=<i>n</i>/<i>m</i> has been developed in connection with the geometrical theory. We consider two different ways for using multiscale bagging of classifiers. The first usage is to construct a confidence set of class labels, instead of a single label. The second usage is to find inputs close to decision boundaries in the context of query by bagging for active learning. It turned out, interestingly, that an appropriate choice of <i>m</i> is <i>m</i>=-<i>n</i>, i.e., σ<sup>2</sup>=-1, for the first usage, and <i>m</i>=∞ , i.e., σ<sup>2</sup>=0, for the second usage.

収録刊行物

  • IEICE transactions on information and systems

    IEICE transactions on information and systems 94(10), 1924-1932, 2011-10-01

    一般社団法人 電子情報通信学会

参考文献:  12件中 1-12件 を表示

各種コード

  • NII論文ID(NAID)
    10030193311
  • NII書誌ID(NCID)
    AA10826272
  • 本文言語コード
    ENG
  • 資料種別
    ART
  • ISSN
    09168532
  • データ提供元
    CJP書誌  J-STAGE 
ページトップへ