PCA Dimensionality Reduction Demonstration

Introduction

• A program for feature reduction, principal component analysis ( PCA ) is offered
• The Main file illustrates the example of how PCA can reduce the number of features using benchmark data-set
• I also demo how to plot the first three components for PCA

Input

• feat : feature vector ( Instances x Features )
• label : label vector ( Instances x 1 )

Output

• new_feat : new features ( Instances x Features )

Example

% Benchmark data set 
load iris.mat

% Set number of principal compoments  
num_pc   = 3; 
% Principal Component Analysis
new_feat = jpca(feat,num_pc);

% Plot first three principal components
jplot(new_feat,label);
legend({'setosa','versicolor','virginica'}); 
xlabel('PC 1'); ylabel('PC 2'); zlabel('PC 3');