GraphLab: k-means info

[Verónica Vergara Larrea (SCC13 Science Lead)]

A team asked about the meaning of the input for k-means. Here is the App Specialists response:

SVD reduces the dimensionality of the input matrix into vectors for each graph node. For example, instead of having 4 million dimensions you need to cluster, you can simply cluster 10 dimensions.  After the clustering you find pairs of nodes with similar singular vectors (means they are similar in some sense).