Multiple-prototype classifier design

Bezdek, James C.; Reichherzer, Thomas; Lim, G.; Attikiouzel, Y.

doi:10.1109/5326.661091

Cited by 103 publications

(51 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The clustering algorithm is applied to generate new prototypes to replace the original training samples, which significantly reduces the number of references in 1-NN classification [53]. Each prototype is a representative of a group of training samples.…”

Section: K-means Based Prototype Learningmentioning

confidence: 99%

A K-Means-Based Multi-Prototype High-Speed Learning System with FPGA-Implemented Coprocessor for 1-NN Searching

Koide

Mattausch

2012

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYIn this paper, we propose a hardware solution for overcoming the problem of high computational demands in a nearest neighbor (NN) based multi-prototype learning system. The multiple prototypes are obtained by a high-speed K-means clustering algorithm utilizing a concept of software-hardware cooperation that takes advantage of the flexibility of the software and the efficiency of the hardware. The one nearest neighbor (1-NN) classifier is used to recognize an object by searching for the nearest Euclidean distance among the prototypes. The major deficiency in conventional implementations for both K-means and 1-NN is the high computational demand of the nearest neighbor searching. This deficiency is resolved by an FPGA-implemented coprocessor that is a VLSI circuit for searching the nearest Euclidean distance. The coprocessor requires 12.9% logic elements and 58% block memory bits of an Altera Stratix III E110 FPGA device. The hardware communicates with the software by a PCI Express (×4) local-bus-compatible interface. We benchmark our learning system against the popular case of handwritten digit recognition in which abundant previous works for comparison are available. In the case of the MNIST database, we could attain the most efficient accuracy rate of 97.91% with 930 prototypes, the learning speed of 1.3 × 10 −4 s/sample and the classification speed of 3.94 × 10 −8 s/character.

show abstract

Section: K-means Based Prototype Learningmentioning

confidence: 99%

A K-Means-Based Multi-Prototype High-Speed Learning System with FPGA-Implemented Coprocessor for 1-NN Searching

Koide

Mattausch

2012

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

show abstract

“…This too is arguable. However, when we mention the instantiations of these two, we specifically consider the LVQ scheme in which the number of 5 In this context, it is also fitting to include a modified Chang's method proposed by Bezdek [14]. 6 Our earlier paper [19] had presented a Hybrid scheme.…”

Section: A Taxonomy Of Prsmentioning

confidence: 99%

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

“…The idea of reducing the computational effort by reducing the number of training samples is known as the prototype approach (for example, see [2] [15], [17], and [34]). According to [34], the training samples used as prototypes may not only be selected but also modified to optimize the classification performance.…”

Section: Comparative Studymentioning

confidence: 99%

“…According to [34], the training samples used as prototypes may not only be selected but also modified to optimize the classification performance. In addition, the training examples may be replaced by multiple prototypes [2]. The methods for prototype generation include competitive learning [2], genetic algorithms [17], and generalized delta rules [15].…”

Section: Comparative Studymentioning

confidence: 99%

See 1 more Smart Citation

A Feature-Based Serial Approach to Classifier Combination

Bunke

Kandel

2002

Pattern Anal Appl

View full text Add to dashboard Cite

Originality and ContributionThe main motivation behind existing methods of classifier combination is improvement of classification accuracy. In this paper we are concerned with a different problem, which is improvement of computational efficiency. Reducing computational effort is extremely important in data-intensive applications of pattern recognition such as clickstream analysis on the web. The classification rate of "lazy" learning methods, like k-NN Classifier, is especially sensitive to the dimensionality of the training set. The computational complexity of a k-NN Classifier depends on two main factors: the number of distance computations (proportional to the number of training examples considered for finding the nearest neighbor of a new instance) and the effort associated with calculating a distance to each training example (proportional to the number of features).Several algorithms have been suggested to reduce the number of distance computations. To the best of our knowledge, this work presents the first attempt to reduce the computational complexity To sum-up, this work is original in at least the following aspects: it uses a feature-based approach to reducing the computational burden of an instance-based classifier; it proposes a serial combination of two k-NN classifiers; and it presents a new feature selection method ("IFN + Relief"). Evaluation results demonstrate the broad practical implications of the proposed methodology.

show abstract

Multiple-prototype classifier design

Cited by 103 publications

References 13 publications

A K-Means-Based Multi-Prototype High-Speed Learning System with FPGA-Implemented Coprocessor for 1-NN Searching

A K-Means-Based Multi-Prototype High-Speed Learning System with FPGA-Implemented Coprocessor for 1-NN Searching

A brief taxonomy and ranking of creative prototype reduction schemes

A Feature-Based Serial Approach to Classifier Combination

Contact Info

Product

Resources

About