Separable linear discriminant analysis

Zhao, Jianhua; Yu, Philip L. H.; Shi, Lei; Li, Shulan

doi:10.1016/j.csda.2012.04.003

Cited by 14 publications

(17 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This paper focuses on the feature extractor design; hence a simple classifier, i.e., minimum-mean-distance is utilized to facilitate the comparison of different feature extractors. We adopt the commonly used non-iterative 2DLDA (N2DLDA) approach [11,7,10] to transform the data into an uncorrelated feature space. The objective of this work is to theoretically derive a discriminance score that can be used for global ranking and hence proper vectorization of the feature matrix components.…”

Section: Problem Definitionmentioning

confidence: 99%

“…In the N2DLDA method [11,7,10], given the matrixvariate data X m×n , the mean of the data in each class and the overall mean are found as…”

Section: Prior Workmentioning

confidence: 99%

“…In these applications, a bidirectional feature extractor such as the commonly used two-directional linear discriminant analysis (2DLDA) solutions [4,5,6,7,8,9,10,11] can be used. These methods require lower computational complexity for training and utilize the inherent matrix-variate structure of the data [11]. For facial images, 2DLDA exploits the correlations between neighboring pixels in both column and row directions.…”

Section: Introductionmentioning

confidence: 99%

“…For simplicity, this paper focuses on non-iterative 2DLDA [11,7,10], although the results can be extended to iterative 2DLDA solutions [4,8] in the future. This paper follows with the definition of the classification problem in Section 2, review of related literature in Section 3, derivation of the proposed method in Section 4, experimental analysis in Section 5, and final comments and future directions in Section 6.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Ranking 2DLDA features based on fisher discriminance

Mahanta

Plataniotis

2014

2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

In classification of matrix-variate data, two-directional linear discriminant analysis (2DLDA) methods extract discriminant features while preserving and utilizing the matrix structure. These methods provide computational efficiency and improved performance in small sample size problems. Existing 2DLDA solutions produce a feature matrix which is commonly vectorized for processing by conventional vectorbased classifiers. However, the vectorization step requires a one-dimensional ranking of features according to their discriminance power. We first demonstrate that independent column-wise and row-wise ranking provided by 2DLDA is not sufficient for uniquely sorting the resulting features, and does not guarantee the selection of the most discriminant features. Then, we theoretically derive the desired global ranking score based on Fisher's criterion. The current results focus on non-iterative solutions, but future extensions to iterative 2DLDA variants are possible. Face recognition experiments using images from the PIE data set are used to demonstrate the theoretically proved improvements over the existing solutions.

show abstract

Section: Problem Definitionmentioning

confidence: 99%

“…In the N2DLDA method [11,7,10], given the matrixvariate data X m×n , the mean of the data in each class and the overall mean are found as…”

Section: Prior Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Ranking 2DLDA features based on fisher discriminance

Mahanta

Plataniotis

2014

2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…it does not take into account class labels), the Linear Discriminant Analysis (LDA) is a popular supervised technique which is widely used in computer-vision, pattern recognition, machine learning and other related fields [6]. LDA performs an optimal projection by maximizing the distance between classes and minimizing the distance between samples within each class at the same time [7].…”

Section: Linear Discriminant Analysismentioning

confidence: 99%