Elsevier

Pattern Recognition

Volume 29, Issue 4, April 1996, Pages 581-588
Pattern Recognition

Construction of class regions by a randomized algorithm: a randomized subclass method

https://doi.org/10.1016/0031-3203(95)00107-7Get rights and content

Abstract

A randomized algorithm is proposed for solving the problem of finding hyper-rectangles, sufficiently approximating the true region in each class. This method yields a suboptimal solution, but is more efficient than previous methods. The performance is analysed based on a criterion of PAC (Probably Approximately Correct) learning. Experimental results show that the proposed method can solve large problems which were not able to be solved previously.

References (12)

  • M. Kudo et al.

    Feature selection based on the structural indices of categories

    Pattern Recognition

    (1993)
  • D. Lubell

    A short proof of sperner's lemma

    J. Comb. Theory

    (1966)
  • M. Ichino

    A nonparametric multiclass pattern classifier

    IEEE Trans. Syst. Man Cybernet.

    (1979)
  • M. Kudo et al.

    Optimal subclasses with dichotomous variables for feature selection and discrimination

    IEEE Trans. Syst. Man Cybernet.

    (1989)
  • M. Ichino

    Optimum feature selection for decision functions

    IEICE

    (1989)
  • M. Kudo et al.

    Analysis of the structure of classes and its applications-subclass approach

There are more references available in the full text version of this article.

Cited by (26)

  • An efficient construction and application usefulness of rectangle greedy covers

    2014, Pattern Recognition
    Citation Excerpt :

    They proposed a method that uses all the MEPSs [4], but the proposed MEPS enumeration algorithm was too slow to use practically. Later, they proposed the randomized subclass method (RSM) [2], which uses k randomly generated MEPSs for each class, where k is an input parameter. An RGC needs more time to construct than a set of rectangles constructed by RSM but uses a smaller number of rectangles in most cases without degrading classification performance so much.

  • Convex sets as prototypes for classifying patterns

    2009, Engineering Applications of Artificial Intelligence
    Citation Excerpt :

    We presented the preliminary version of this work in Takigawa et al. (2005), focusing on minimum enclosing balls as convex sets. The idea was originated in the special case of minimum enclosing axis-parallel boxes (Kudo et al., 1996; Takigawa et al., 2004). It is also related with logical analysis of data (LAD) with boxes (Alexe and Hammer, 2006; Eckstein et al., 2002), ball-based combinatorial classifier (Cannon et al., 2002; Cannon and Cowen, 2004; Priebe et al., 2003; Marchette, 2004; DeVinney, 2003), conventional prototype-based methods such as nearest neighbors and learning vector quantization (Hastie et al., 2001), and classifiers based on explicit computational geometric structure such as support vector machine (Schölkopf and Smola, 2001).

View all citing articles on Scopus
View full text