Main APIs

Main APIs

maximise_function

opt_val, opt_pt, history = maximise_function(func, domain, max_capital,
                                             capital_type='num_evals',
                                             opt_method='bo',
                                             config=None,
                                             options=None,
                                             reporter='default')

Maximises a function func over the domain domain.

Arguments:

func: The function to be maximised.
domain: The domain over which the function should be maximised, should be an instance of the Domain class in exd/domains.py. If domain is a list of the form [[l1, u1], [l2, u2], ...] where li < ui, then we create a Euclidean domain with lower bounds li and upper bounds ui along each dimension.
max_capital: The maximum capital (time budget or number of evaluations) available for optimisation.
opt_method: The method used for optimisation. Could be one of 'bo', 'rand', 'ga', 'ea', 'direct', or 'pdoo'. Default is 'bo'. 'bo': Bayesian optimisation, 'ea'/'ga': Evolutionary algorithm, 'rand': Random search, 'direct': Dividing Rectangles, 'pdoo': PDOO.
worker_manager: The type of worker manager. Should be an instance of exd.worker_manager.AbstractWorkerManager or a string with one of the following values: 'default', 'synthetic', 'multiprocessing', 'scheduling'.
num_workers: The number of parallel workers (i.e. number of evaluations to carry out in parallel).
capital_type: The type of capital. Should be one of 'num_evals', 'return_value' or 'realtime'. Default is 'num_evals' which indicates the number of evaluations. If 'realtime', we will use wall clock time.
config: Either a configuration file or or parameters returned by exd.cp_domain_utils.load_config_file. config can be None only if domain is a EuclideanDomain object.
options: Additional hyper-parameters for optimisation, as a namespace.
reporter: A stream to print progress made during optimisation, or one of the following strings 'default', 'silent'. If 'silent', then it suppresses all outputs. If 'default', writes to stdout.

Alternatively, domain and fidelity space could be None if config is either a path_name to a configuration file or has configuration parameters.

Returns:

opt_val: The maximum value found during the optimisation procedure.
opt_pt: The corresponding optimum point.
history: A record of the optimisation procedure which include the point evaluated and the values at each time step.

See here, here, here, here, here, here, here, and here for more detailed examples.

minimise_function

opt_val, opt_pt, history = minimise_function(func, domain, max_capital,
                                             capital_type='num_evals',
                                             opt_method='bo',
                                             config=None,
                                             options=None,
                                             reporter='default')

Arguments:
Same as maximise_function (see above), but now func is to be minimised.
Returns:
Same as maximise_function (see above), but now opt_val is the minimum value found during optimisation.

maximise_multifidelity_function

opt_val, opt_pt, history = maximise_multifidelity_function(func, fidel_space, domain,
                             fidel_to_opt, fidel_cost_func, max_capital,
                             capital_type='return_value',
                             opt_method='bo',
                             config=None,
                             options=None,
                             reporter='default'):

Maximises a multi-fidelity function func over the domain domain and fidelity space 'fidel_space'. See the BOCA paper for more information on multi-fidelity optimisation.

Arguments:

'func': The function to be maximised. Takes two arguments func(z, x) where z is a member of the fidelity space and x is a member of the domain.
'fidel_space': The fidelity space from which the approximations are obtained. Should be an instance of the Domain class in exd/domains.py. If of the form [[l1, u1], [l2, u2], ...] where li < ui, then we will create a Euclidean domain with lower bounds li and upper bounds ui along each dimension.
'domain': The domain over which the function should be maximised, should be an instance of the Domain class in exd/domains.py. If domain is a list of the form [[l1, u1], [l2, u2], ...] where li < ui, then we will create a Euclidean domain with lower bounds li and upper bounds ui along each dimension.
'fidel_to_opt': The point at the fidelity space at which we wish to maximise func.
'max_capital': The maximum capital (time budget or number of evaluations) available for optimisation.
'capital_type': The type of capital. Should be one of 'return_value' or 'realtime'. Default is 'return_value' which indicates we will use the value returned by fidel_cost_func. If 'realtime', we will use wall clock time.
'opt_method': The method used for optimisation. Could be one of 'bo' or 'rand'. Default is 'bo'. 'bo': Bayesian optimisation, 'rand': Random search.
'config': Either a configuration file or or parameters returned by exd.cp_domain_utils.load_config_file. config can be None only if domain is a EuclideanDomain object.
'options': Additional hyper-parameters for optimisation, as a namespace.
'reporter': A stream to print progress made during optimisation, or one of the following strings 'default', 'silent'. If 'silent', then it suppresses all outputs. If 'default', writes to stdout.

Alternatively, domain and fidelity space could be None if config is either a path_name to a configuration file or has configuration parameters.

Returns:

'opt_val': The maximum value found during the optimisation procdure.
'opt_pt': The corresponding optimum point.
'history': A record of the optimisation procedure which include the point evaluated and the values at each time step.

See here, here, here, here, here, and here for more detailed examples.

minimise_multifidelity_function

opt_val, opt_pt, history = minimise_multifidelity_function(func, fidel_space, domain,
                             fidel_to_opt, fidel_cost_func, max_capital,
                             capital_type='return_value',
                             opt_method='bo',
                             config=None,
                             options=None,
                             reporter='default'):

Arguments:
Same as maximise_multifidelity_function (see above), but now func is to be minimised.
Returns:
Same as maximise_multifidelity_function (see above), but now opt_val is the minimum value found during optimisation.

multiobjective_maximise_function

pareto_values, pareto_points, history = multiobjective_maximise_functions(
                                          funcs, domain, max_capital,
                                          capital_type='num_evals',
                                          opt_method='bo',
                                          config=None,
                                          options=None,
                                          reporter='default')

Jointly optimises the functions funcs over the domain domain.

Arguments:

funcs: The functions to be jointly maximised.
domain: The domain over which the function should be maximised, should be an instance of the Domain class in exd/domains.py. If domain is a list of the form [[l1, u1], [l2, u2], ...] where li < ui, then we create a Euclidean domain with lower bounds li and upper bounds ui along each dimension.
max_capital: The maximum capital (time budget or number of evaluations) available for optimisation.
capital_type: The type of capital. Should be one of 'num_evals', 'return_value' or 'realtime'. Default is 'num_evals' which indicates the number of evaluations. If 'realtime', we will use wall clock time.
'opt_method': The method used for optimisation. Could be one of 'bo' or 'rand'. Default is 'bo'. 'bo': Bayesian optimisation, 'rand': Random search.
'config': Either a configuration file or or parameters returned by exd.cp_domain_utils.load_config_file. config can be None only if domain is a EuclideanDomain object.
'options': Additional hyper-parameters for optimisation, as a namespace.
'reporter': A stream to print progress made during optimisation, or one of the following strings 'default', 'silent'. If 'silent', then it suppresses all outputs. If 'default', writes to stdout.

Alternatively, domain and fidelity space could be None if config is either a path_name to a configuration file or has configuration parameters.

Returns:

'pareto_values': The pareto optimal values found during the optimisation procdure.
'pareto_points': The corresponding pareto optimum points in the domain.
'history': A record of the optimisation procedure which include the point evaluated and the values at each time step.

See here and here for examples.

multiobjective_minimise_function

pareto_values, pareto_points, history = multiobjective_minimise_functions(
                                          funcs, domain, max_capital,
                                          capital_type='num_evals',
                                          opt_method='bo',
                                          config=None,
                                          options=None,
                                          reporter='default')

Arguments:
Same as multiobjective_maximise_functions (see above), but now funcs are to be minimised.
Returns:
Same as multiobjective_maximise_functions (see above), but now pareto_values are the Pareto optimal minimum values found during optimisation.

See here and here for examples.

maximize_function

Alternative spelling for maximise_function.

minimize_function

Alternative spelling for minimise_function.

maximize_multifidelity_function

Alternative spelling for maximise_function.

minimize_multifidelity_function

Alternative spelling for minimise_multifidelity_function.

multiobjective_maximize_functions

Alternative spelling for multiobjective_maximise_functions.

multiobjective_minimize_functions

Alternative spelling for multiobjective_minimise_functions.