propulate.propagators.base

Module Contents

Classes

Propagator

Initialize a propagator with given parameters.

Stochastic

Initialize a stochastic propagator that is only applied with a specified probability.

Conditional

Initialize a conditional propagator.

Compose

Initialize a composed propagator.

SelectMin

Initialize an elitist selection propagator.

SelectMax

Initialize an anti-elitist propagator.

SelectUniform

Initialize a random-selection propagator.

InitUniform

Initialize a random-initialization propagator.

Gaussian

Initialize Gaussian propagator.

Functions

_check_compatible(→ bool)

Check compatibility of two propagators for stacking them together sequentially with Compose.
propulate.propagators.base._check_compatible(out1: int, in2: int) → bool

Check compatibility of two propagators for stacking them together sequentially with Compose.

Parameters:

  • out1 (int) – The number of output individuals returned by the first propagator.

  • in2 (int) – The number of input individuals taken by the second propagator.

Returns:

True if the input propagators can be stacked: False if not.

Return type:

bool

class propulate.propagators.base.Propagator(parents: int = 0, offspring: int = 0, rng: random.Random | None = None)

Initialize a propagator with given parameters.

Parameters:

  • parents (int, optional) – The number of input individuals (-1 for any). Default is 0 for abstract base class.

  • offspring (int, optional) – The number of output individuals to breed. Default is 0 for abstract base class.

  • rng (random.Random, optional) – The separate random number generator for the Propulate optimization.

Raises:

ValueError – If the number of offspring to breed is zero.

abstract __call__(inds: List[propulate.population.Individual]) → List[propulate.population.Individual] | propulate.population.Individual

Apply the propagator (not implemented for abstract base class).

Parameters:

inds (List[propulate.population.Individual]) – The input individuals the propagator is applied to.

Returns:

The individual(s) bred by applying the propagator. While this abstract base class method actually returns None, each concrete child class of Propagator should return an Individual instance or a list of them.

Return type:

List[propulate.population.Individual] | propulate.population.Individual

Raises:

NotImplementedError – Whenever called (abstract base class method).

class propulate.propagators.base.Stochastic(parents: int = 0, offspring: int = 0, probability: float = 1.0, rng: random.Random | None = random.Random())

Bases: Propagator

Initialize a stochastic propagator that is only applied with a specified probability.

Parameters:

  • parents (int, optional) – The number of input individuals (-1 for any). Default is 0.

  • offspring (int, optional) – The number of output individuals. Default is 0.

  • probability (float, optional) – The probability of application. Default is 0.0.

  • rng (random.Random, optional) – The separate random number generator for the Propulate optimization.

Raises:

ValueError – If the number of offspring to breed is zero.

class propulate.propagators.base.Conditional(pop_size: int, true_prop: Propagator, false_prop: Propagator, parents: int = -1, offspring: int = -1)

Bases: Propagator

Initialize a conditional propagator.

Parameters:

  • pop_size (int) – The breeding population size.

  • true_prop (propulate.propagators.Propagator) – The propagator applied if the current population’s size equals at least pop_size.

  • false_prop (propulate.propagators.Propagator) – The propagator applied if the current population’s size is less than pop_size.

  • parents (int, optional) – The number of input individuals (-1 for any). Default is -1.

  • offspring (int) – The number of output individuals to breed. Default is -1.

__call__(inds: List[propulate.population.Individual]) → List[propulate.population.Individual] | propulate.population.Individual

Apply conditional propagator.

Parameters:

inds (List[propulate.population.Individual]) – The input individuals the propagator is applied to.

Returns:

The output individuals returned by the conditional propagator.

Return type:

List[propulate.population.Individual]

class propulate.propagators.base.Compose(propagators: List[Propagator])

Bases: Propagator

Initialize a composed propagator.

Parameters:

propagators (List[propulate.propagators.Propagator]) – The propagators to be stacked together sequentially.

Raises:

ValueError – If the propagators to stack are incompatible in terms of number of input and output individuals.

__call__(inds: List[propulate.population.Individual]) → List[propulate.population.Individual] | propulate.population.Individual

Apply the composed propagator.

Parameters:

inds (List[propulate.population.Individual]) – The input individuals the propagator is applied to.

Returns:

The output individuals after application of the propagator.

Return type:

List[propulate.population.Individual]

class propulate.propagators.base.SelectMin(offspring: int)

Bases: Propagator

Initialize an elitist selection propagator.

Parameters:

offspring (int) – The number of offspring (individuals to be selected).

__call__(inds: List[propulate.population.Individual]) → List[propulate.population.Individual]

Apply the elitist-selection propagator.

Parameters:

inds (List[propulate.population.Individual]) – The input individuals the propagator is applied to.

Returns:

The selected output individuals after application of the propagator.

Return type:

List[propulate.population.Individual]

Raises:

ValueError – If more individuals than put in shall be selected.

class propulate.propagators.base.SelectMax(offspring: int)

Bases: Propagator

Initialize an anti-elitist propagator.

Parameters:

offspring (int) – The number of offspring (individuals to be selected).

__call__(inds: List[propulate.population.Individual]) → List[propulate.population.Individual]

Apply the anti-elitist-selection propagator.

Parameters:

inds (List[propulate.population.Individual]) – The individuals the propagator is applied to.

Returns:

The selected individuals after application of the propagator.

Return type:

List[propulate.population.Individual]

Raises:

ValueError – If more individuals than put in shall be selected.

class propulate.propagators.base.SelectUniform(offspring: int, rng: random.Random | None = None)

Bases: Propagator

Initialize a random-selection propagator.

Parameters:

  • offspring (int) – The number of offspring (individuals to be selected).

  • rng (random.Random, optional) – The separate random number generator for the Propulate optimization.

__call__(inds: List[propulate.population.Individual]) → List[propulate.population.Individual]

Apply the uniform-selection propagator.

Parameters:

inds (List[propulate.population.Individual]) – The individuals the propagator is applied to.

Returns:

The selected individuals after application of the propagator.

Return type:

List[propulate.population.Individual]

Raises:

ValueError – If more individuals than put in shall be selected.

class propulate.propagators.base.InitUniform(limits: Mapping[str, Tuple[float, float] | Tuple[int, int] | Tuple[str, Ellipsis]], parents: int = 0, probability: float = 1.0, rng: random.Random | None = random.Random())

Bases: Stochastic

Initialize a random-initialization propagator.

Parameters:

  • limits (Dict[str, Tuple[float, float]] | Dict[str, Tuple[int, int]] | Dict[str, Tuple[str, ...]]) – The search space, i.e., the limits of (hyper-)parameters to be optimized.

  • parents (int, optional) – The number of parents. Default is 0.

  • probability (float, optional) – The probability of application. Default is 1.0.

  • rng (random.Random, optional) – The separate random number generator for the Propulate optimization.

__call__(*inds: propulate.population.Individual) → propulate.population.Individual

Apply the uniform-initialization propagator.

Parameters:

inds (propulate.population.Individual) – The individuals the propagator is applied to.

Returns:

The output individual after application of the propagator.

Return type:

propulate.population.Individual

Raises:

ValueError – If a parameter’s type is invalid, i.e., not float (continuous), int (ordinal), or str (categorical).

class propulate.propagators.base.Gaussian(limits: Dict[str, Tuple[float, float]], scale: float, rng: numpy.random.Generator)

Bases: Propagator

Initialize Gaussian propagator.

Parameters:

  • limits (Dict[str, Tuple[float, float]] | Dict[str, Tuple[int, int]] | Dict[str, Tuple[str, ...]]) – The search space, i.e., limits of (hyper-)parameters to be optimized.

  • scale (float) – The standard deviation of the Gaussian distribution.

  • rng (random.Random) – The separate random number generator for the Propulate optimization.

__call__(inds: List[propulate.population.Individual]) → propulate.population.Individual

Apply the Gaussian propagator.

Parameters:

inds (propulate.population.Individual) – The individuals the propagator is applied to.

Returns:

The output individual after application of the propagator.

Return type:

propulate.population.Individual