R/soft_coord.R
soft_coord.Rd
Runs an iterative coordinate descent algorithm to minimize the SURE for mode-specific soft thresholding estimators.
soft_coord(c_obj, lambda_init, c_init, itermax = 1000, tol = 10^-4, print_iter = TRUE, tau2 = 1, use_sure = TRUE)
c_obj | The output from |
---|---|
lambda_init | A vector of numerics of length \(n\). The initial starting values for the thresholding parameters. |
c_init | A positive numeric. The starting value of the scaling parameter. |
itermax | A positive integer. The maximum number of Newton steps to iterate through. |
tol | A positive numeric. The stopping criterion. |
print_iter | A logical. Should we print the updates of the Newton Step? |
tau2 | A positive numeric. The variance. Assumed known and defaults to 1. |
use_sure | A logical. Which stopping criterion should we use? The mean
absolute difference in the parameters ( |
c
A numeric. The final value of the scaling parameter.
lambda
A vector of numerics. The final values of the thresholding
parameters.
est
An array of numerics. The final mean estimate.
Gerard, D., & Hoff, P. (2015). Adaptive Higher-order Spectral Estimators. arXiv preprint arXiv:1505.02114.