NPVecchia-package: Large Scale Covariance Estimation

In katzfuss-group/NPvecchia: Large Scale Covariance Estimation

Large Scale Covariance Estimation

Description

A nearest-neighbor approach similar to Vecchia's is employed to quickly scale covariance estimation to high dimensions (with few samples). Only using the nearest neighbors ensures sparsity in the Cholesky of the precision matrix, while our Bayesian approach adds further regularization to the non-zero elements.

Details

The main functions are listed below (and have C++ versions by adding "_c" to the name). See the vignette for further details.

minus_loglikeli gives the integrated log-likelihood as a function of 3 hyperparameters

thetas_to_priors gives a List of the priors from the hyperparameters

get_posts gets the posteriors from these priors

samp_posts gets the MAP from the posteriors to give an estimate

Author(s)

Brian Kidd

Maintainer: Brian Kidd <[e-mail]>

References

Eventually, this will include a link to the Arxiv paper.

katzfuss-group/NPvecchia documentation built on April 15, 2022, 2:23 a.m.