fista.nuclear function

A helper function for implementing FISTA algorithm to estimate low-rank matrix

Function to estimate low-rank matrix using FISTA algorithm

fista.nuclear(A, b, lambda, d, niter, backtracking = TRUE, phi.true)

Arguments

• A: A n by p design matrix
• b: A correspond vector, or a matrix
• lambda: tuning parameter
• d: model dimension
• niter: the maximum number of iterations required for applying FISTA algorithm
• backtracking: a boolean argument, indicate whether use backtracking or not
• phi.true: true model parameter, only available for simulations

Returns

A list object, including

• phi.hat: Estimated low-rank matrix
• obj.vals: Values of objective function for all iterations
• rel.err: Relative error to the true model parameter, only available for simulation