For very large data sets we might save memory and computing time at minimal accuracy loss by using 'tapered' versions of the covariances as described here.
Tapering makes the covariance matrix sparse so it can be stored in smaller memory and inverted with optimized methods as in the Eigen library described here.
For very large data sets we might save memory and computing time at minimal accuracy loss by using 'tapered' versions of the covariances as described here.
Tapering makes the covariance matrix sparse so it can be stored in smaller memory and inverted with optimized methods as in the
Eigen
library described here.