%autosave 10
Autosaving every 10 seconds
Code and IPython Notebook
!!AI see presenter notes.
!!AI see notes. requires solid linear algebra knowledge (maybe add references).
Looking at smallest eigenvectors, whereas PCA looks at largest eigenvectors, different problem.
Unnormalised spectral clustering tries to balance the size of the clusters
Normalised spectral clustering tries to balance the degree (density) of the clusters.