slides - SNAP - Stanford University
slides - SNAP - Stanford University
slides - SNAP - Stanford University
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Approximates pp the optimal p cut [ [Shi‐Malik, , ’00] ]<br />
Can be used to approximate the optimal k‐way normalized cut<br />
Emphasizes cohesive clusters<br />
Increases the h unevenness in i the h distribution di ib i of f the h data d<br />
Associations between similar points are amplified, associations<br />
between dissimilar points are attenuated<br />
Th The data dt bbegins i t to “ “approximate i t a clustering” l t i ”<br />
Well‐separated space<br />
Transforms data to a new “embedded embedded space”, space ,<br />
consisting of k orthogonal basis vectors<br />
NB: Multiple eigenvectors prevent instability due to<br />
information f loss l<br />
11/8/2010 Jure Leskovec, <strong>Stanford</strong> CS224W: Social and Information Network Analysis, http://cs224w.stanford.edu 42