particle_swarm_optimizer

A flexible Particle Swarm Optimization implementation in Python. Created as a personal project by Julian Pryde, not intended for professional use.

Photos

Quickstart

1. Pull repository
2. Update forcing_function file with your forcing function
3. Update arguments file with your parameters
4. Run main.py
5. Output will be to STDOUT

Features

• Handles problems in any number of dimensions
• Stops when either no particle has moved by more than a specific distance or when the best particle has not changed for 100 iterations
• Outputs the total number of groups of particles that have gathered together around local maxima/minima

Parameters

• num_particles: Sets the number of particles in the simulation
• function: <max/min> Tell the program whether to find a maximum or a minimum of the forcing function
• local_radius: The radius over which each particle will search for other better particles when determining its velocity for the next iteration
• velocity_coefficient: Universal scaling factor to determine the velocity of each particle
• starting_sigma: Coefficient for determining the gaussian spread of each particle on the first iteration
• exit_criterion: If no particle moves less than this number in the normalized axes, the program will assume it has reached a max/min and stop.
• annealing_lifetime: The sigma value will go down incrementally until this iteration number.

Dependencies:

• decimal
• math
• random
• matplotlib