amarszalek / pygamoo

GAme theory based framework for Multiobjective Optimization in Python

pyAGAMO

Asynchronous GAme theory based framework for MultiObjective Optimization in Python

Installation

The package can be download and install form github usig pip:

pip install git+https://github.com/amarszalek/pyAGAMO.git#egg=pyagamo

Quick example (Jupyter Notebook: Quick_example.ipynb)

The described below example demonstrates how to set up and run a multi-objective optimization using the pyAGAMO framework. It includes parameter definitions, initialization of objectives and players, running the optimization process (all on single computing machine), and visualizing the results. This example provides a clear starting point for users to understand and implement their own optimization problems using pyAGAMO.

Import Necessary Modules

This imports essential modules from the pyAGAMO framework and matplotlib for visualization.

# problem
from pyagamo.objectives import RE36
# player
from pyagamo.players import ClonalSelection
# manager
from pyagamo import AGAMO
import matplotlib.pyplot as plt

Set Main Parameters of the Algorithm

Defines key parameters:

• max_eval: Maximum number of objective function evaluations.
• npop: Population size.
• change_iter: Frequency of changes assigned decision variables to players (number of iterations).
• next_iter: Maximum of optimization steps that can be performed during the remaining player's iteration (-1 means no limitation)
• max_front: Maximum number of solutions in the result population.

max_eval = 10000
npop = 25
change_iter = 1
next_iter = -1
max_front = 100

Set Parameters for the Clonal Selection Algorithm

Defines parameters specific to the Clonal Selection algorithm:

• nclone: Number of clones.
• mutate_args: Hypermutation arguments.
• sup: Suppression factor.

player_parm = {"nclone": 15, "mutate_args": [0.45, 0.9, 0.01], 'sup': 0.0}

Create Objectives and Players

Creates a list of objectives and a list of players. Each objective and player is instantiated with specific parameters.

objs = [RE36(i, obj=i+1) for i in range(3)]
players = [ClonalSelection(i, npop, player_parm) for i in range(3)]

Initialize and Start the AGAMO Manager

Initializes the AGAMO manager with defined parameters, adds objectives and players, and starts the optimization process.

agamo = AGAMO(max_eval, change_iter, next_iter, max_front)
agamo.add_objectives(objs)
agamo.add_players(players)
agamo.init()
agamo.start_optimize()

Get and Visualize Results

Retrieves the results of the optimization process, closes the AGAMO manager, and visualizes the Pareto front of the optimization results using a 3D scatter plot.

res = agamo.get_results()
agamo.close()

ax = plt.figure('Front Pareto', figsize=(6, 6)).add_subplot(111, projection='3d')
front_eval = res['front_eval']
ax.scatter(front_eval[:, 0], front_eval[:, 1], front_eval[:, 2], marker='o', label=str(max_eval))
ax.grid(True)
ax.view_init(30, 30)
plt.legend()
plt.show()

Usage examples

• Quick example Quick_example.ipynb, simple example of usage, all objects on single computing machine.

Documentation

pyAGAMO's documentation can be found at https://pyagamo.readthedocs.io.