Content deleted Content added
→Algorithms: move maths from lead, avoid confusion of two Ss |
→Relationship with k-means: punctuation for refs |
||
Line 45:
\max_G \text{ trace }\left(G^TG\right).
</math>
This problem is equivalent to the spectral clustering problem when the identity constraints on <math>F</math> are relaxed. In particular, the weighted kernel ''k''-means problem can be reformulated as a spectral clustering (graph partitioning) problem and vice-versa. The output of the algorithms are eigenvectors which do not satisfy the identity requirements for indicator variables defined by <math>F</math>. Hence, post-processing of the eigenvectors is required for the equivalence between the problems.<ref name="dhillon2004kernel">{{cite conference
| author = Dhillon, I.S. and Guan, Y. and Kulis, B.
| year = 2004
Line 52:
| pages = 551--556
| organization = ACM
}}</ref>
Transforming the spectral clustering problem into a weighted kernel ''k''-means problem greatly reduces the computational burden.
== References ==
|