A bootstrap technique for nearest neighbor classifier design

Hamamoto, Yoshihiko; Uchimura, Shunji; Tomita, Shunsuke

doi:10.1109/34.566814

Cited by 105 publications

(72 citation statements)

References 19 publications

(1 reference statement)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…To overcome the difficulty with the sparsity of data in such spaces, researchers have investigated using a decision threshold [1] [4] and enhancing the training set by bootstrapping [5].…”

Section: Introductionmentioning

confidence: 99%

Nearest neighbors in random subspaces

Ho¹

1998

Advances in Pattern Recognition

155

View full text Add to dashboard Cite

Abstract. Recent studies have shown that the random subspace method can be used to create multiple independent tree-classifiers that can be combined to improve accuracy. We apply the procedure to k-nearestneighbor classifiers and show that it can achieve similar results. We examine the effects of several parameters of the method by experiments using data from a digit recognition problem. We show that the combined accuracies follow a trend of increase with increasing number of component classifiers, and that with an appropriate subspace dimensionality, the method can be superior to simple k-nearest-neighbor classification. The method's superiority is maintained when smaller number of training prototypes are available, i.e., when conventional knn classifiers suffer most heavily from the curse of dimensionality.

show abstract

“…To overcome the difficulty with the sparsity of data in such spaces, researchers have investigated using a decision threshold [1] [4] and enhancing the training set by bootstrapping [5].…”

Section: Introductionmentioning

confidence: 99%

Nearest neighbors in random subspaces

Ho¹

1998

Advances in Pattern Recognition

155

View full text Add to dashboard Cite

show abstract

“…The schemes reported by Chang [5] (the PNN) 5 , Xie et al [10] (the VQ), Hamamoto et al [11] (the BT), Song and Lee [13] (the LVQ) and Kim and Oommen [19] (HYB), create new prototype vectors (and do not merely select training samples) in such a way that these prototypes represent all the vectors in the original set in the "best" possible manner. The methods of the Hart [3] (the CNN), Gates [4] (the RNN), Ritter et al [6] (the SNN), Tomek [7] (the mCNN), Devijer and Kittler [8] (the ENN), Fukunaga [9] (the PZN), Bezdek [15] (the GA and RS), Sanchez et al [16] (the PG), and Lipowezky [17] (the PF) are those in which the prototype vectors are merely selected.…”

Section: A Taxonomy Of Prsmentioning

confidence: 99%

“…In other cases, we cannot estimate, a priori the maximum number of iterations which should be done. The former is the criterion employed in the CNN [3], RNN [4], SNN [6], mCNN [7], ENN [8], VQ [10], BT [11] and GA (RS) [15], and the latter is the criterion used in the PNN [5], PZN [9], LVQ [13], the PG [16] and HYB [19] 6 .…”

Section: A Taxonomy Of Prsmentioning

confidence: 99%

“…In the approaches of the VQ [10] and BT [11], we can decide the number of prototype vectors based on the control strategy of the algorithms. On the other hand, in the case of the CNN [3], RNN [4], PNN [5], SNN [6], mCNN [7], ENN [8], PZN [9], GA (RS) [15], PG [16], PF [17] and HYB [19], the number of prototype vectors is itself a random variable, and we cannot determine it until the processes are finished.…”

Section: A Taxonomy Of Prsmentioning

confidence: 99%

See 1 more Smart Citation

A brief taxonomy and ranking of creative prototype reduction schemes

Kim

Oommen

2003

Pattern Analysis & Applications

View full text Add to dashboard Cite

Various Prototype Reduction Schemes (PRS) have been reported in the literature. Based on their operating characteristics, these schemes fall into two fairly distinct categories -those which are of a creative sort, and those which are essentially selective. The norms for evaluating these methods are typically, the reduction rate and the classification accuracy. It is generally believed that the former class of methods is superior to the latter.In this paper, we report the results of executing various creative PRS, and attempt to comparatively quantify their capabilities. The paper presents a brief taxonomy of the various reported PRS schemes. Our experimental results for three artificial data sets, and for samples involving real-life data sets, demonstrate that no single method is uniformly superior to the others for all kinds of applications. This result, though consistent with the findings of [2], is, in one sense, counter-intuitive, because the various researchers have presented their specific PRS with the hope that it would be superior to the previously reported methods. But the fact is that while one method is superior in certain domains, it is inferior to another method when dealing with a data set with markedly different characteristics. The conclusion of this study is that the question of determining when one method is superior to another remains open. Indeed, it appears as if the designers of the Pattern Recognition system will have to choose the appropriate PRS based to the specific characteristics of the data that they are studying. The paper also suggests answers to various hypotheses that relate to the accuracies and reduction rates of families of PRS.

show abstract

“…• Selection of a design subset of prototypes from a given set of prototype vectors [5,8,11,18,21] and generation of prototype reference vectors [7].…”

Section: Related Workmentioning

confidence: 99%

Efficient Implementation of Nearest Neighbor Classification

Herrero

Navarro

Advances in Soft Computing

View full text Add to dashboard Cite

A bootstrap technique for nearest neighbor classifier design

Cited by 105 publications

References 19 publications

Nearest neighbors in random subspaces

Nearest neighbors in random subspaces

A brief taxonomy and ranking of creative prototype reduction schemes

Efficient Implementation of Nearest Neighbor Classification

Contact Info

Product

Resources

About