PSO

class globe.optimizers.particles.PSO.PSO(bounds, n_particles=200, iter=1000, dt=0.1, beta=100000.0, alpha=1, batch_size=0, filter_type=None, verbose=False)[source]

Bases: Particles_Optimizer

Interface for the social only PSO optimizer with optional particle filtering.

Parameters:

  • bounds (ndarray) – The bounds of the search space.

  • n_particles (int) – The number of particles.

  • iter (int) – The number of iterations.

  • dt (float) – The time step.

  • beta (float) – The inverse temperature for using a Gibbs measure to select the global best instead of the argmin. Default is 0 (no Gibbs measure).

  • alpha (float) – The coefficient to decrease the step size.

  • batch_size (int) – The batch size for the mini-batch optimization. If 0, no mini-batch optimization is used.

  • filter_type (str or None, optional) – The type of filter to apply to particles: - None: No filtering (default) - “quantile”: Filters out particles judged as non-relevant based on quantile

  • verbose (bool) – Whether to print information about the optimization process.