SLDR

A MATLAB toolbox for supervised linear dimension reduction (SLDR) including LDA, HLDA, MMDA, WHMMDA, PLS-DA, and SDA

Codes for the following papers were implemented:

1. Heteroscedastic Max–Min distance analysis for dimensionality reduction (WHMMDA)
2. Heteroscedastic max-min distance analysis (WHMMDA)
3. Max-min distance analysis by using sequential SDP relaxation for dimension reduction (MMDA)
4. Linear dimensionality reduction via a heteroscedastic extension of LDA: the Chernoff criterion (HLDA)
5. Multiclass partial least squares discriminant analysis: Taking the right way—A critical tutorial (PLS-DA)
6. Stochastic discriminant analysis for linear supervised dimension reduction (SDA)

Note:

To avoid matrix singularity in computations, we employ Marchenko–Pastur for denoising covariance matrices.

1. Introduction.

This package includes the prototype MATLAB codes for supervised linear dimension reduction (SLDR).

The implemented methods include:

1. Linear discriminant analysis (LDA)
2. Heteroscedastic extension of LDA (HLDA)
3. Max-min distance analysis (MMDA)
4. Heteroscedastic extension of MMDA (WHMMDA)
5. Partial least squares discriminant analysis (PLS‐DA)
6. Stochastic discriminant analysis (SDA)

2. Usage & Dependency.

Dependency:

 If you want to use MMDA or WHMMDA, you should download the following zip file & extract it in the "cvx-toolbox" or current directory
 CVX MATLAB toolbox for Windows can be downloaded from [website](http://web.cvxr.com/cvx/cvx-w64.zip)

Usage:

Run and check "demo_run_methods.m" and you'll see the below results for all methods