This repository contains the source codes for the numerical experiments described in the article:
Y. Bello-Cruz, J.G. Melo, L.F. Prudente, and R.V.G. Serra,
A Proximal Gradient Method with an Explicit Linesearch for Multiobjective Optimization, technical report, 2024.
Date: November 2024
The repository includes the following routines:
main.m
Controls the execution of the experiments, allowing users to:- Select the problem to be solved.
- Choose the optimization algorithm to apply.
-
ProxGrad.m
Contains implementations of:- Explicit-MPG: Explicit Multiobjective Proximal Gradient algorithm.
- Armijo-MPG: Multiobjective Proximal Gradient with Armijo line search.
- Implicit-MPG: Implicit Multiobjective Proximal Gradient algorithm.
-
ProxGradAcc.m
Contains implementations of:- Accelerated-MPG: Accelerated Multiobjective Proximal Gradient algorithm.
- Normal-MPG: Normal Proximal Gradient algorithm.
-
inip.m
Defines the initial parameters for the optimization problems. -
evalg.m
Implements the functions G_j. -
evalgradg.m
Computes the gradient of the functions G_j. -
evalh.m
Implements the functions H_j. -
datas.m
Generates the data for the functions H_j. -
subproblemProxGrad.m
Solves the subproblem required in the proximal gradient methods.
-
explicitLS.m
Implements the explicit line search used in Explicit-MPG. -
explicitLSone.m
Handles the line search for a single objective function in Explicit-MPG. -
armijo.m
Implements the Armijo line search used in Armijo-MPG. -
implicitLS.m
Implements the implicit line search used in Implicit-MPG.
To run the numerical experiments:
- Open the file
main.m. - Select the desired problem and algorithm in the code.
- Run the script and view the results on the screen.
This project is licensed under the GNU General Public License (GPL) version 3. For more information, see the LICENSE file.