GPyOpt.optimization package

Submodules

GPyOpt.optimization.acquisition_optimizer module

class GPyOpt.optimization.acquisition_optimizer.AcquisitionOptimizer(space, optimizer='lbfgs', **kwargs)

Bases: object

General class for acquisition optimizers defined in domains with mix of discrete, continuous, bandit variables

Parameters:
  • space – design space class from GPyOpt.
  • optimizer – optimizer to use. Can be selected among: - ‘lbfgs’: L-BFGS. - ‘DIRECT’: Dividing Rectangles. - ‘CMA’: covariance matrix adaptation.
optimize(f=None, df=None, f_df=None, duplicate_manager=None)

Optimizes the input function.

Parameters:
  • f – function to optimize.
  • df – gradient of the function to optimize.
  • f_df – returns both the function to optimize and its gradient.
class GPyOpt.optimization.acquisition_optimizer.ContextManager(space, context=None)

Bases: object

class to handle the context variable in the optimizer :param space: design space class from GPyOpt. :param context: dictionary of variables and their contex values

GPyOpt.optimization.anchor_points_generator module

class GPyOpt.optimization.anchor_points_generator.AnchorPointsGenerator(space, design_type, num_samples)

Bases: object

get(num_anchor=5, duplicate_manager=None, unique=False, context_manager=None)
get_anchor_point_scores(X)
class GPyOpt.optimization.anchor_points_generator.ObjectiveAnchorPointsGenerator(space, design_type, objective, num_samples=1000)

Bases: GPyOpt.optimization.anchor_points_generator.AnchorPointsGenerator

get_anchor_point_scores(X)
class GPyOpt.optimization.anchor_points_generator.RandomAnchorPointsGenerator(space, design_type, num_samples=10000)

Bases: GPyOpt.optimization.anchor_points_generator.AnchorPointsGenerator

get_anchor_point_scores(X)
class GPyOpt.optimization.anchor_points_generator.ThompsonSamplingAnchorPointsGenerator(space, design_type, model, num_samples=25000)

Bases: GPyOpt.optimization.anchor_points_generator.AnchorPointsGenerator

get_anchor_point_scores(X)

GPyOpt.optimization.optimizer module

class GPyOpt.optimization.optimizer.OptCma(bounds, maxiter=1000)

Bases: GPyOpt.optimization.optimizer.Optimizer

Wrapper the Covariance Matrix Adaptation Evolutionary strategy (CMA-ES) optimization method. It works generating an stochastic search based on multivariate Gaussian samples. Only requires f and the box constraints to work.

optimize(x0, f=None, df=None, f_df=None)
Parameters:
  • x0 – initial point for a local optimizer.
  • f – function to optimize.
  • df – gradient of the function to optimize.
  • f_df – returns both the function to optimize and its gradient.
class GPyOpt.optimization.optimizer.OptDirect(bounds, maxiter=1000)

Bases: GPyOpt.optimization.optimizer.Optimizer

Wrapper for DIRECT optimization method. It works partitioning iteratively the domain of the function. Only requires f and the box constraints to work.

optimize(x0, f=None, df=None, f_df=None)
Parameters:
  • x0 – initial point for a local optimizer.
  • f – function to optimize.
  • df – gradient of the function to optimize.
  • f_df – returns both the function to optimize and its gradient.
class GPyOpt.optimization.optimizer.OptLbfgs(bounds, maxiter=1000)

Bases: GPyOpt.optimization.optimizer.Optimizer

Wrapper for l-bfgs-b to use the true or the approximate gradients.

optimize(x0, f=None, df=None, f_df=None)
Parameters:
  • x0 – initial point for a local optimizer.
  • f – function to optimize.
  • df – gradient of the function to optimize.
  • f_df – returns both the function to optimize and its gradient.
class GPyOpt.optimization.optimizer.OptimizationWithContext(x0, f, df=None, f_df=None, context_manager=None)

Bases: object

df_nc(x)

Wrapper of the derivative of f: takes an input x with size of the not fixed dimensions expands it and evaluates the gradient of the entire function.

f_df_nc(x)

Wrapper of the derivative of f: takes an input x with size of the not fixed dimensions expands it and evaluates the gradient of the entire function.

f_nc(x)

Wrapper of f: takes an input x with size of the noncontext dimensions expands it and evaluates the entire function.

class GPyOpt.optimization.optimizer.Optimizer(bounds)

Bases: object

Class for a general acquisition optimizer.

Parameters:bounds – list of tuple with bounds of the optimizer
optimize(x0, f=None, df=None, f_df=None)
Parameters:
  • x0 – initial point for a local optimizer.
  • f – function to optimize.
  • df – gradient of the function to optimize.
  • f_df – returns both the function to optimize and its gradient.
GPyOpt.optimization.optimizer.apply_optimizer(optimizer, x0, f=None, df=None, f_df=None, duplicate_manager=None, context_manager=None, space=None)
Parameters:
  • x0 – initial point for a local optimizer (x0 can be defined with or without the context included).
  • f – function to optimize.
  • df – gradient of the function to optimize.
  • f_df – returns both the function to optimize and its gradient.
  • duplicate_manager – logic to check for duplicate (always operates in the full space, context included)
  • context_manager – If provided, x0 (and the optimizer) operates in the space without the context
  • space – GPyOpt class design space.
GPyOpt.optimization.optimizer.choose_optimizer(optimizer_name, bounds)

Selects the type of local optimizer

Module contents