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.

SRDA: an Efficient Algorithm for Large-Scale

SRDA: an Efficient Algorithm for Large-Scale Discriminant Analysis

Training Linear Discriminant Analysis in Linear Time

Efficient Kernel Discriminant Analysis via Spectral Regression

Speed up kernel discriminant analysis

Spectral Regression for Efficient Regularized Subspace Learning

Robust sparse manifold discriminant analysis

Linear discriminant analysis with spectral regularization

Spectral regression: a regression framework for efficient regularized subspace learning

Understanding and Evaluating Sparse Linear Discriminant Analysis

Quadratic Discriminant Analysis by Projection

Sparse Exponential Discriminant Analysis

Sparse tensor discriminant analysis

A Two-Stage Linear Discriminant Analysis Via QR-decomposition.

High-Dimensional Regularized Discriminant Analysis

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

Local Feature Discriminant Projection

3E-LDA

Constrained Linear Discrimination Analysis and Face Recognition

A Direct Estimation Approach to Sparse Linear Discriminant Analysis

Direct linear discriminant analysis based on column pivoting QR decomposition and economic SVD