-
[show abstract]
[hide abstract]
ABSTRACT: Principal component analysis (PCA) has been widely applied in the area of computer science. It is well-known that PCA is a popular transform method and the transform result is not directly related to a sole feature component of the original sample. However, in this paper, we try to apply principal components analysis (PCA) to feature selection. The proposed method well addresses the feature selection issue, from a viewpoint of numerical analysis. The analysis clearly shows that PCA has the potential to perform feature selection and is able to select a number of important individuals from all the feature components. Our method assumes that different feature components of original samples have different effects on feature extraction result and exploits the eigenvectors of the covariance matrix of PCA to evaluate the significance of each feature component of the original sample. When evaluating the significance of the feature components, the proposed method takes a number of eigenvectors into account. Then it uses a reasonable scheme to perform feature selection. The devised algorithm is not only subject to the nature of PCA but also computationally efficient. The experimental results on face recognition show that when the proposed method is able to greatly reduce the dimensionality of the original samples, it also does not bring the decrease in the recognition accuracy.
System Science, Engineering Design and Manufacturing Informatization (ICSEM), 2010 International Conference on; 12/2010
-
[show abstract]
[hide abstract]
ABSTRACT: In this paper we propose a novel feature selection method based on linear discriminant analysis (LDA). To view feature selection as a numerical computation problem, the paper shows, for the first time, that it is feasible to employ LDA for feature selection. The proposed method also shows that different components statistically have different effects on the feature selection result, which can be evaluated by the components of the eigenvector. As there are multiple eigenvectors, the proposed method takes a small number of eigenvectors into account when evaluating the effect of the component of the sample data. The experimental results on face recognition show that the proposed method is not only able to greatly reduce the dimensionality of the original samples, but also able to yield promising classification accuracies.
Intelligent System Design and Engineering Application (ISDEA), 2010 International Conference on; 11/2010
-
2009 Second International Symposium on Computational Intelligence and Design, ISCID 2009, Changsha, Hunan, China, 12-14 December 2009, 2 Volumes; 01/2009
-
Neurocomputing. 01/2008; 71:1883-1888.
-
[show abstract]
[hide abstract]
ABSTRACT: Maximum scatter difference (MSD) discriminant criterion was a recently presented binary discriminant criterion for pattern classification that utilizes the generalized scatter difference rather than the generalized Rayleigh quotient as a class separability measure, thereby avoiding the singularity problem when addressing small-sample-size problems. MSD classifiers based on this criterion have been quite effective on face-recognition tasks, but as they are binary classifiers, they are not as efficient on large-scale classification tasks. To address the problem, this paper generalizes the classification-oriented binary criterion to its multiple counterpart--multiple MSD (MMSD) discriminant criterion for facial feature extraction. The MMSD feature-extraction method, which is based on this novel discriminant criterion, is a new subspace-based feature-extraction method. Unlike most other subspace-based feature-extraction methods, the MMSD computes its discriminant vectors from both the range of the between-class scatter matrix and the null space of the within-class scatter matrix. The MMSD is theoretically elegant and easy to calculate. Extensive experimental studies conducted on the benchmark database, FERET, show that the MMSD out-performs state-of-the-art facial feature-extraction methods such as null space method, direct linear discriminant analysis (LDA), eigenface, Fisherface, and complete LDA.
IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics) 01/2008; 37(6):1599-606. · 3.08 Impact Factor
-
IEEE Transactions on Systems, Man, and Cybernetics, Part B. 01/2007; 37:1599-1606.
-
IJPRAI. 01/2007; 21:1085-1101.
-
Neurocomputing. 01/2007; 71:191-196.
-
Pattern Anal. Appl. 01/2007; 10:165-174.
-
[show abstract]
[hide abstract]
ABSTRACT: Image recognition on streaming data is one of the most challenging topics in Image and Video Technology and incremental dimensionality
reduction algorithms play a key role in online image recognition. In this paper, we present a novel supervised dimensionality
reduction algorithm—Incremental Weighted Karhunen-Loève expansion based on the Between-class scatter matrix (IWKLB) for image
recognition on streaming data. In comparison with Incremental PCA, IWKLB is more effective in terms of recognition rate. In
comparison with Incremental LDA, it is free of small sample size problems and can directly be applied to high-dimensional
image spaces with high efficiency. Experimental results conducted on AR, one benchmark face image database, demonstrate that
IWKLB is more effective than IPCA and ILDA.
KeywordsDimensionality reduction-supervised learning-image recognition-streaming data-incremental algorithm
12/2006: pages 198-207;
-
Advances in Image and Video Technology, First Pacific Rim Symposium, PSIVT 2006, Hsinchu, Taiwan, December 10-13, 2006, Proceedings; 01/2006
-
Neurocomputing. 01/2006; 69:1683-1687.
-
[show abstract]
[hide abstract]
ABSTRACT: It is well known that the classification effectiveness of the text categorization system is not simply a matter of learning algorithms. Text representation factors are also at work. This paper will consider the ways in which the effectiveness of text classifiers is linked to the five text representation factors: stop words removal, word stemming, indexing, weighting, and normalization. Statistical analyses of experimental results show that performing normalization can always promote effectiveness of text classifiers significantly. The effects of the other factors are not as great as expected. Contradictory to common sense, a simple binary indexing method can sometimes be helpful for text categorization.
Formal Pattern Analysis & Applications 01/2005; 8(1):199-209. · 0.74 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: The performance of a novel binary linear classifier named as minimum norm minimum squared-error (MNMSE), which is based on a refined minimum squared-error discriminant criterion is evaluated in this paper. Experimental results show that MNMSE is very effective and efficient for many pattern recognition problems. In most cases it can compete with support vector machines in recognition rate and be more efficient than the methods.
Control, Automation, Robotics and Vision Conference, 2004. ICARCV 2004 8th; 01/2005
-
Pattern Anal. Appl. 01/2005; 8:199-209.
-
Pattern Recognition. 01/2005; 38:311-313.
-
Pattern Recognition. 01/2004; 37:1953-1955.
-
[show abstract]
[hide abstract]
ABSTRACT: Kernel principal component analysis (KPCA) extracts features of samples with an efficiency in inverse proportion to the size of the training sample set. In this paper, we develop a novel method to improve KPCA-based feature extraction. The developed method is the first one that is methodologically consistent with KPCA. Experiments on several benchmark datasets illustrate that the feature extraction process derived from the novel method is much more efficient than that associated with KPCA. Moreover, the classification accuracy generated from the developed method is similar to that of KPCA.
Neurocomputing.
-
[show abstract]
[hide abstract]
ABSTRACT: To overcome shortcomings of traditional linear discriminant analysis such as failing to extract features of data with complex distributions, a new LDA approach is proposed in this paper. This approach is based on the following perspective: for a sample, the sample that is from the same class and is the farthest away from this sample, is typical intra-class sample of this sample. On the other hand, for the same sample, the nearest neighbor from each of other classes is called typical inter-class sample. In practice "typical samples" of a sample have indicative meaning for the space relation between this sample and the others. The new LDA approach bases definitions of between-class and within-class scatter matrices on these typical samples. As a result, the linear transform associated with our approach is able to maximize the distances between a sample and the corresponding typical inter-class samples, while minimizing the distance between the same sample and the typical intra-class sample. The proposed new approach is able to extract features of not only data with simple distributions but also the data with complex distributions, which means that the new LDA approach has wider applicability than traditional LDA.
-
[show abstract]
[hide abstract]
ABSTRACT: Maximum scatter difference (MSD) discriminant criterion was a recently presented binary discriminant criterion for pattern classification that utilizes the generalized scatter difference rather than the generalized Rayleigh quotient as a class separability measure, thereby avoiding the singularity problem when addressing small-sample-size problems. MSD classifiers based on this criterion have been quite effective on face-recognition tasks, but as they are binary classifiers, they are not as efficient on large-scale classification tasks. To address the problem, this paper generalizes the classification-oriented binary criterion to its multiple counterpart—multiple MSD (MMSD) discriminant criterion for facial feature extraction. The MMSD feature- extraction method, which is based on this novel discriminant criterion, is a new subspace-based feature-extraction method. Unlike most other subspace-based feature-extraction methods, the MMSD computes its discriminant vectors from both the range of the between-class scatter matrix and the null space of the within-class scatter matrix. The MMSD is theoretically elegant and easy to calculate. Extensive experimental studies conducted on the benchmark database, FERET, show that the MMSD outperforms state-of-the-art facial feature-extraction methods such as null space method, direct linear discriminant analysis (LDA), eigenface, Fisherface, and complete LDA.
