Abstract:Linear discriminant analysis (LDA) represents a simple yet powerful technique for partitioning a p-dimensional feature vector into one of K classes based on a linear projection learned from N labeled observations. However, it is well-established that in the high-dimensional setting (p > N) the underlying projection estimator degenerates. Moreover, any linear discriminate function involving a large number of features may be difficult to interpret. To ameliorate these issues, two general categories of sparse LDA modifications have been proposed, both to reduce the number of active features and to stabilize the resulting projections. The first, based on optimal scoring, is more straightforward to implement and analyze but has been heavily criticized for its ambiguous connection with the original LDA formulation. In contrast, a second strategy applies sparse penalty functions directly to the original LDA objective but requires additional heuristic trade-off parameters, has unknown global and local minima properties, and requires a greedy sequential optimization procedure. In all cases the choice of sparse regularizer can be important, but no rigorous guidelines have been provided regarding which penalty might be preferable. Against this backdrop, we winnow down the broad space of candidate sparse LDA algorithms and promote a specific selection based on optimal scoring coupled with a particular, complementary sparse regularizer. This overall process ultimately progresses our understanding of sparse LDA in general, while leading to targeted modifications of existing algorithms that produce superior results in practice on three high-dimensional gene data sets.

Linear Dimensionality Reduction Using Relevance Weighted LDA

Computational and Theoretical Analysis of Supervised Dimensionality Reduction

3E-LDA

Adaptive Local Linear Discriminant Analysis

A New Reduced-Rank Linear Discriminant Analysis Method and Its Applications

Linear Sequence Discriminant Analysis: A Model-Based Dimensionality Reduction Method for Vector Sequences

Sr-LDA:Sparse and Reduced-Rank Linear Discriminant Analysis for High Dimensional Matrix

Optimal Dimensionality Discriminant Analysis and Its Application to Image Recognition

Optimal Feature Selection for Sparse Linear Discriminant Analysis and Its Applications in Gene Expression Data.

Multimodal Linear Discriminant Analysis Via Structural Sparsity

Face Recognition Using Novel LDA-Based Algorithms.

Dimension reduction of multimodal data by auto-weighted local discriminant analysis

Discriminative Dimensionality Reduction for Multi-Dimensional Sequences.

Linear Discriminant Regularized Regression

Disambiguation Enabled Linear Discriminant Analysis for Partial Label Dimensionality Reduction

Understanding and Evaluating Sparse Linear Discriminant Analysis

Sparse Local Discriminant Projections for Feature Extraction

Saliency-based Weighted Multi-label Linear Discriminant Analysis

Supervised Linear Dimension-Reduction Methods: Review, Extensions, and Comparisons

Locality-regularized Linear Regression Discriminant Analysis for Feature Extraction

Nonlinear Dimensionality Reduction for Discriminative Analytics of Multiple Datasets