Abstract:Dimensionality reduction is a crucial step for pattern recognition and data mining tasks to overcome the curse of dimensionality. Principal component analysis (PCA) is a traditional technique for unsupervised dimensionality reduction, which is often employed to seek a projection to best represent the data in a least-squares sense, but if the original data is nonlinear structure, the performance of PCA will quickly drop. An supervised dimensionality reduction algorithm called Linear discriminant analysis (LDA) seeks for an embedding transformation, which can work well with Gaussian distribution data or single-modal data, but for non-Gaussian distribution data or multimodal data, it gives undesired results. What is worse, the dimension of LDA cannot be more than the number of classes. In order to solve these issues, Local shrunk discriminant analysis (LSDA) is proposed in this work to process the non-Gaussian distribution data or multimodal data, which not only incorporate both the linear and nonlinear structures of original data, but also learn the pattern shrinking to make the data more flexible to fit the manifold structure. Further, LSDA has more strong generalization performance, whose objective function will become local LDA and traditional LDA when different extreme parameters are utilized respectively. What is more, a new efficient optimization algorithm is introduced to solve the non-convex objective function with low computational cost. Compared with other related approaches, such as PCA, LDA and local LDA, the proposed method can derive a subspace which is more suitable for non-Gaussian distribution and real data. Promising experimental results on different kinds of data sets demonstrate the effectiveness of the proposed approach.

Generalized Re-Weighting Local Sampling Mean Discriminant Analysis

Sample Weighting: an Inherent Approach for Outlier Suppressing Discriminant Analysis

Generalized LDA Using Relevance Weighting and Evolution Strategy.

Self-weighted Learning Framework for Adaptive Locality Discriminant Analysis

Local Feature Discriminant Projection

Weighted Linear Discriminant Analysis: An Effective Feature Extraction Method for Multi-Class Imbalanced Datasets

By Normalizing To Improve Generalized Foley-Sammon Transform In High-Dimensional Spaces-----With Application To Face Recognition

Self-Weighted Adaptive Locality Discriminant Analysis

Local Sampling Mean Discriminant Analysis with Kernels

Local Shrunk Discriminant Analysis (LSDA).

Locality-regularized Linear Regression Discriminant Analysis for Feature Extraction

Locality Adaptive Discriminant Analysis Framework

Local and Weighted Maximum Margin Discriminant Analysis

Adaptive and Fuzzy Locality Discriminant Analysis for Dimensionality Reduction

Linear Dimensionality Reduction Using Relevance Weighted LDA

Generalizing relevance weighted LDA

Feature Extraction By Locality-Based Linear Discriminant Analysis

An Improved Random Sampling LDA for Face Recognition

Self-weighted Robust LDA for Multiclass Classification with Edge Classes

Robust sparse manifold discriminant analysis

3E-LDA