PCA Guide

Specification

Location	Author	Maintained by
isl/math/functions.py	Kenny Erleben	DIKU

Principal Component Analysis

This tutorial shows how to use our module’s principal component analysis PCA. This implementation finds the eigenvalues using the NumPy LinAlg eig library.

Examples

First import the module

    import isl.math.functions as func

Example 1

Then define a dataset having the signature [[x0,y0,z0], … , [xN,yN,zN]]

    import numpy as np
    M = 10
    P = np.random.rand(M, 3)
    mean, principal_componets, eigenvectors = func.PCA(P)
    mean, principal_componets, eigenvectors

Output

    (array([0.36885196, 0.54575938, 0.46111808]),
     array([0.04772988, 0.11529473, 0.102832  ]),
     array([[ 0.8278388 ,  0.47242563, -0.30248461],
        [ 0.55871669, -0.74262082,  0.36925598],
        [ 0.05018538,  0.47468763,  0.87872241]]))

Example 2

Find the most describing eigenvector

    import numpy as np
    M = 10
    P = np.random.rand(M, 3)
    eigenvector = func.direction_of_most_variance(P)
    eigenvector

Output

    array([-0.65208111, -0.62503103,  0.42909955])