Skip to main content

NSMA: A Memetic Procedure for Global Multi-Objective Optimization.

Project description

Python 3.10 license DOI

Alt Text

NSMA: A Memetic Procedure for Global Multi-Objective Optimization

Implementation of the NSMA Algorithm proposed in

Lapucci, M., Mansueto, P. & Schoen, F. A memetic procedure for global multi-objective optimization. Math. Prog. Comp. (2022).

If you have used our code for research purposes, please cite the publication mentioned above. For the sake of simplicity, we provide the Bibtex format:

@Article{Lapucci2022,
    author={Lapucci, Matteo and Mansueto, Pierluigi and Schoen, Fabio},
    title={A memetic procedure for global multi-objective optimization},
    journal={Mathematical Programming Computation},
    year={2022},
    month={Nov},
    day={22},
    issn={1867-2957},
    doi={10.1007/s12532-022-00231-3},
    url={https://doi.org/10.1007/s12532-022-00231-3}
}

Main Dependencies

  • python v3.9
  • pip v22.2.2
  • numpy v1.22.3
  • scipy v1.7.3
  • matplotlib Windows: v3.5.3, Linux: v3.5.2, MacOSX: v3.6.1
  • tensorflow Windows: v2.9.1, Linux: v2.8.2, MacOSX: v2.10.0
  • gurobipy v9.5.2
  • progressbar2 v4.2.0

Gurobi Optimizer

In order to run some parts of the code, the Gurobi Optimizer needs to be installed and, in addition, a valid Gurobi licence is required. However, the employment of the Gurobi Optimizer is not mandatory to execute the code. Indeed, we provide alternative scripts where the HiGHS dual simplex solver implementation by SciPy is used.

Usage

We refer to the code documentation and the related GitHub repository for all the information.

An usage example could be the following:

import tensorflow as tf
from nsma.algorithms.memetic.nsma import NSMA
from nsma.problems.man.man_instance import MAN1
from nsma.general_utils.pareto_utils import points_initialization

tf.compat.v1.disable_eager_execution()

session = tf.compat.v1.Session()
with session.as_default():
    algorithm = NSMA(max_iter=None, 
                     max_time=2, 
                     max_f_evals=None, 
                     verbose=True, 
                     verbose_interspace=10, 
                     plot_pareto_front=True, 
                     plot_pareto_solutions=False, 
                     plot_dpi=100, 
                     pop_size=100, 
                     crossover_probability=0.9, 
                     crossover_eta=20, 
                     mutation_eta=20, 
                     shift=10, 
                     crowding_quantile=0.9, 
                     n_opt=5, 
                     FMOPG_max_iter=5, 
                     theta_for_stationarity=-1e-10, 
                     theta_tol=-1e-1, 
                     theta_dec_factor=10**(-0.5), 
                     gurobi=True, 
                     gurobi_method=1, 
                     gurobi_verbose=False, 
                     ALS_alpha_0=1, 
                     ALS_delta=0.5, 
                     ALS_beta=10**-4, 
                     ALS_min_alpha=1e-7)
    problem = MAN1(n=5)
    initial_p_list, initial_f_list, n_initial_points = points_initialization(problem, 'hyper', 5)
    p_list, f_list, elapsed_time = algorithm.search(initial_p_list, initial_f_list, problem)

Contact

If you have any question, feel free to contact me:

Pierluigi Mansueto
Global Optimization Laboratory (GOL)
University of Florence
Email: pierluigi dot mansueto at unifi dot it

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

nsma-1.0.13.tar.gz (37.9 kB view hashes)

Uploaded Source

Built Distribution

nsma-1.0.13-py3-none-any.whl (56.8 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page