ActiveSetPursuit.jl

Sparse least-squares solver
Author MPF-Optimization-Laboratory
Popularity
0 Stars
Updated Last
3 Months Ago
Started In
June 2024

ActiveSetPursuit

Stable Dev Build Status

ActiveSetPursuit

ActiveSetPursuit is a Julia package providing a framework for solving several variations of the sparse optimization problem:

$$\underset{x}{\text{minimize}} \hspace{0.5em} \lambda ||x||_1 + \frac{1}{2} ||Ax - b||^2_2$$

Implemented Algorithms:

  • Basis Pursuit Denoising
  • Orthogonal Matching Pursuit
  • Homotopy Basis Pursuit Denoising

Installation:

The package can be installed as follows:

julia> ] add ActiveSetPursuit

Example Usage

The asp_bpdn function within the ActiveSetPursuit package efficiently solves the basis pursuit denoising problem.

Ensure that the following inputs are defined before running the asp_bpdn function:

  • A: The matrix or linear operator, size m-by-n.
  • b: The vector of observations or measurements, size m.
  • λin: A nonnegative scalar that serves as the regularization parameter.

To solve the basis pursuit denoising problem, execute the following command in Julia:

N, M = 1000, 300
A = randn(N, M)
xref = zeros(M)
xref[randperm(M)[1:20]] = randn(20)  # 20 non-zero entries
b = A * xref
const λin = 0.0
tracer = asp_bpdn(A, b, λin, traceFlag = true)

After the optimization process completes, if traceFlag was set to true, the solution vector and the regularization parameter at any iteration itn can be accessed as follows:

xx, λ = tracer[itn]

To extract the final iterate:

x_final, λ_final = tracer[end]

Used By Packages

No packages found.