Chunk #6 — METHODS — LOW-DIMENSIONAL EMBEDDING BY EIGEN-ANALYSIS

Source

Discovering genetic ancestry using spectral graph theory.

Embedded

yes

Text

Since we are using the low-dimensional embeddings for clustering it is important to know what the distances in the embedding space represent. An analysis shows that the kernel H induces a natural Euclidean distance m(i,j) between individuals: (1)m(i,j)2=L−1‖yi−yj‖2=hii+hjj−2hij,i,j=1,…,N.