pyhelios.ParameterOptimization Namespace Reference

Classes
class	Adam
	AdamW gradient-based optimization. More...
class	BayesianOptimization
	Bayesian optimization with a Gaussian process surrogate. More...
class	BLXAlphaCrossover
	Component-wise blend crossover. More...
class	BLXPCACrossover
	Blend crossover in PCA-transformed space, for non-separable problems. More...
class	BOBYQA
	BOBYQA derivative-free local optimization. More...
class	CMAES
	Covariance Matrix Adaptation Evolution Strategy. More...
class	ConstrainedResult
	One evaluation of a constrained simulation. More...
class	GeneticAlgorithm
	Genetic algorithm settings. More...
class	HybridMutation
	Mixture of PCA-Gaussian, PCA-Cauchy, and random-direction mutation. More...
class	IsotropicMutation
	Isotropic Gaussian mutation applied to all genes together. More...
class	LBFGS
	L-BFGS gradient-based optimization. More...
class	OptimizationResult
	Outcome of an optimization run. More...
class	Parameter
	A single optimizable parameter. More...
class	ParameterOptimization
	Optimize named model parameters against an objective function. More...
class	ParameterOptimizationError
	Exception raised for ParameterOptimization-specific errors. More...
class	ParameterType
	Kind of an optimizable parameter. More...
class	PerGeneMutation
	Per-gene Gaussian mutation. More...
class	SLSQP
	SLSQP gradient-based optimization. More...

Functions
Callable[[Dict[str, float]], ConstrainedResult]	make_constrained_simulation (Callable[[Dict[str, float]], float] objective, Callable[[Dict[str, float]], Dict[str, float]] objective_gradient, Sequence[tuple] constraints)
	Compose separate objective and constraint callables into one simulation.

Variables
	logger = logging.getLogger(__name__)
	CrossoverOperator = Union[BLXAlphaCrossover, BLXPCACrossover]
	MutationOperator = Union[PerGeneMutation, IsotropicMutation, HybridMutation]
	AlgorithmSettings
dict	_ALGORITHM_INFO

Function Documentation

make_constrained_simulation()

Callable[[Dict[str, float]], ConstrainedResult] pyhelios.ParameterOptimization.make_constrained_simulation	(	Callable[[Dict[str, float]], float]	objective,
		Callable[[Dict[str, float]], Dict[str, float]]	objective_gradient,
		Sequence[tuple]	constraints )

Compose separate objective and constraint callables into one simulation.

A convenience for problems whose constraints really are independent functions. Note that it calls every function at each parameter point, so it forfeits the single-pass advantage of writing one combined simulation: if computing the objective and the constraints shares expensive work – as it does when they come from one Helios simulation – write a :class:ConstrainedResult directly instead.

Parameters

objective	Callable receiving {name: value} and returning a scalar cost
objective_gradient	Callable returning {name: partial derivative}
constraints	Sequence of (function, gradient) pairs, each satisfied when function(params) <= 0

Returns

A callable suitable for :meth:ParameterOptimization.runConstrained

Example

>>> simulation = make_constrained_simulation( ... lambda p: p["x"] ** 2, ... lambda p: {"x": 2 * p["x"]}, ... [(lambda p: 1.0 - p["x"], lambda p: {"x": -1.0})])

Definition at line 533 of file ParameterOptimization.py.

Variable Documentation

_ALGORITHM_INFO

dict pyhelios.ParameterOptimization._ALGORITHM_INFO

protected

Definition at line 450 of file ParameterOptimization.py.

AlgorithmSettings

pyhelios.ParameterOptimization.AlgorithmSettings

Initial value:

=  Union[GeneticAlgorithm, BayesianOptimization, CMAES,
                          Adam, LBFGS, BOBYQA, SLSQP]

Definition at line 437 of file ParameterOptimization.py.

CrossoverOperator

pyhelios.ParameterOptimization.CrossoverOperator = Union[BLXAlphaCrossover, BLXPCACrossover]

Definition at line 128 of file ParameterOptimization.py.

logger

pyhelios.ParameterOptimization.logger = logging.getLogger(__name__)

Definition at line 35 of file ParameterOptimization.py.

MutationOperator

pyhelios.ParameterOptimization.MutationOperator = Union[PerGeneMutation, IsotropicMutation, HybridMutation]

Definition at line 129 of file ParameterOptimization.py.