normalize

Normalize the gram matrices so the features lie on the surface of a unit hypersphere, $\hat{k}(x,z)=\frac{k(x,z)}{\sqrt{k(x,x)k(z,z)}}$. When the Gram matrices are not full, i.e., the diagonals do not contain all $k(x,x)$ or $k(z,z)$ entries, the diagonals must be provided in optList.

gram = normalize(gramObj, optList)

gramObj	-	Gram object
optList	-	list of NAME,VALUE option pairs

optList:

'Diag1'	-	Diagonal entries associated with first dimension for $K$ Gram matrices in a $(M\times K)$ matrix.
'Diag2'	-	Diagonal entries associated with second dimension for $K$ Gram matrices in a $(N\times K)$ matrix.