cov.RQ: Rational quadratic covariance function
In JimSkinner/spca: Structured PCA

Rational quadratic covariance function

1	cov.RQ(X, X2, beta, D = NA, ...)

locations = matrix(rnorm(10), ncol=2)
beta      = rnorm(3) # Logarithms of variance, length scale & alpha

K = cov.RQ(locations, beta=beta)
stopifnot(all(Matrix::diag(K)==exp(beta[1]))) # Diagonal is exp(beta[1])
stopifnot(all(svd(K)$d>0)) # K is positive definite
stopifnot(all(K[upper.tri(K)]<K[1,1])) # Largest element is on diagonal