Particle

Represents a single particle in the Particle Swarm Optimization (PSO) algorithm.

Attributes

  • position (numpy.ndarray): The current position of the particle in the search space.

  • velocity (numpy.ndarray): The current velocity of the particle.

  • best_position (numpy.ndarray): The best position discovered by the particle.

  • best_score (float): The objective function value corresponding to the best position.

  • idx (int): An identifier for the particle.

  • iter (int): The current iteration number for the particle.

Methods

  • __init__(self, bounds, init_position='random'): Initializes a new particle with a random or specified initial position and zero initial velocity.

ParticleSwarmOptimizer

Implements the Particle Swarm Optimization (PSO) algorithm for finding the minimum of an objective function in a multi-dimensional space.

Attributes

  • objective_function (callable): The objective function to be minimized.

  • bounds (list of tuples): Defines the search space through lower and upper bounds for each parameter.

  • num_particles (int): The number of particles in the swarm.

  • max_iter (int): The maximum number of iterations for optimization.

  • converged_threshold (float): The convergence threshold for optimization termination.

  • recorder (LossRecorder): An instance of the LossRecorder class to track losses and optimization parameters.

  • iter (int): The current iteration number.

  • options (dict): Configuration options for PSO.

Methods

  • __init__(self, objective_function, bounds, update_boundary_frequency=5, begin_update_boundary_frequency_iter=0, converged_threshold=1e-6, max_no_improvement_iters=8, num_particles=30, max_iter=100, options=None): Initializes the ParticleSwarmOptimizer with objective function, search space bounds, and optimization parameters.

  • update_inertia_weight(self, iter): Adjusts the inertia weight based on the current iteration for exploration and exploitation balance.

  • adaptive_bounds(self, current_iter): Dynamically adjusts the search space bounds based on the global best position for exploration.

  • optimize(self, opt_folder=None): Executes the PSO algorithm to minimize the objective function within defined bounds.

  • optimize_mpi(self, max_processes, opt_folder=None): Executes PSO in parallel using multiple processes for faster optimization.

optimize_mpi is a parallel version of optimize that utilizes multiprocessing to speed up the computation.

optimize executes PSO sequentially, while optimize_mpi executes PSO in parallel, leveraging multiple processes for faster computation. optimize_mpi requires specifying the maximum number of worker processes for parallel computation. optimize_mpi also provides an optional parameter to specify an output folder for storing optimization artifacts.

update_inertia_weight, adaptive_bounds, and optimize methods handle different aspects of the PSO algorithm, such as adjusting inertia weight, dynamically updating search space bounds, and executing the optimization process.

The options parameter in the ParticleSwarmOptimizer class allows users to customize PSO parameters such as inertia weight, acceleration coefficients, etc. options provide flexibility for fine-tuning the PSO algorithm according to specific optimization tasks.

The recorder attribute of the ParticleSwarmOptimizer class is an instance of the LossRecorder class, used for recording losses and optimization parameters over iterations. recorder facilitates monitoring and analyzing optimization progress, aiding in result interpretation and troubleshooting. recorder records losses, global best parameters, and iteration details for comprehensive optimization tracking and analysis.