GCN-Langevin

class globe.optimizers.particles.common_noise.GCN.GCN_Langevin.GCN_Langevin(bounds, n_particles=200, iter=100, dt=0.1, beta=1, filter_type=None, sigma_noise=1, independent_noise=True, verbose=False)[source]

Bases: Particles_Optimizer

Interface for the Geometric Common Noise Langevin 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.

  • 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

  • sigma_noise (float) – The kernel bandwidth for the common noise.

  • independent_noise (bool) – Whether to use independent noise for each particle.

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