fista_sparse {multivar} | R Documentation |
Function for estimating multiple-subjbect Vector Autoregression (VAR) models using Fast Iterative Shrinkage-Thresholding Algorithm (FISTA; Beck and Teboulle, 2009)
fista_sparse(A, b, lambda, x_true, niter, backtrack, w = NULL, conv = 1e-10)
A |
An N x P design matrix. |
b |
An N x P outcome matrix. |
lambda |
Regularization parameter. |
x_true |
Numeric matrix containing the true transition matrix (if available). |
niter |
Integer giving the maximum number of iterations. |
backtrack |
Logical. If backtracking should be used in the FISTA algorithm. |
w |
Numeric matrix containing the weights (if available). |
conv |
Convergance criterion. |
Function Under Development
This is a prototype function and is currently under development.
Fisher, Z.F., Kim, Y., and Pipiras, V. (Under Review) Penalized Estimation and Forecasting of Multiple Subject Intensive Longitudinal Data.
Beck A. and Teboulle, M. (2009). A Fast Iterative Shrinkage-Thresholding Algorithm for Linear Inverse Problems. SIAM J. Img. Sci. 2, 1, 183–202.
theta <- matrix(rnorm(9),3,3) data <- var_sim(20, theta, diag(.1,3)) datalag <- embed(data, 2) b <- datalag[,1:3] A <- datalag[,4:6] fista_sparse(A, b, 1, theta, niter = 1, backtrack = TRUE)