Principal Component Analysis (PCA)
Principal Component Analysis (PCA)
Principal Component Analysis (PCA)
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Maximum Variance Projection<br />
• Consider a random vector x with arbitrary joint pdf<br />
• in <strong>PCA</strong> we assume that x has zero mean<br />
• we can always achieve this by constructing x’=x-µ<br />
• Scatter plot and projection y onto a “weight” vector w<br />
w<br />
y<br />
n<br />
T<br />
= w x = ∑<br />
i=<br />
1<br />
cos(<br />
α )<br />
• Motivating question: y is a scalar random variable that<br />
depends on w; for which w is the variance of y maximal?<br />
Jochen Triesch, UC San Diego, http://cogsci.ucsd.edu/~triesch 2<br />
x<br />
y<br />
=<br />
w<br />
×<br />
x<br />
×<br />
w<br />
i<br />
x<br />
i