Crossover

Collection of crossover operators for evolutionary algorithms.

Includes implementations of arithmetic, blend (BLX-Alpha), simulated binary (SBX), intermediate, heuristic, and differential crossover. Designed for use with real-valued vectors and adaptable to various evolutionary strategies.

evolib.operators.crossover.crossover_arithmetic(parent1_para, parent2_para, num_children=2)

Perform arithmetic crossover between two parent vectors.

Each child’s genes are a weighted average of the corresponding genes of the parents, using a randomly chosen mixing coefficient alpha ∈ [0, 1].

Parameters:

• parent1_para (ndarray) – Parameter vector of the first parent.

• parent2_para (ndarray) – Parameter vector of the second parent.

• num_children (int) – Number of children to return (1 or 2). Default is 2.

Returns:

One or two offspring vectors.

Return type:

Union[ndarray, Tuple[ndarray, ndarray]]

Raises:

ValueError – If parent vectors have different lengths or num_children is invalid.

evolib.operators.crossover.crossover_blend_alpha(parent1_para, parent2_para, alpha=0.5, num_children=2)

Perform Blend-Alpha Crossover (BLX-Alpha) on two parent vectors.

This operator creates one or two offspring by sampling each gene from an extended interval around the parent genes.

Parameters:

• parent1_para (ndarray) – Parameter vector of the first parent.

• parent2_para (ndarray) – Parameter vector of the second parent.

• alpha (float) – Expansion factor for the sampling interval. Default is 0.5.

• num_children (int) – Number of children to generate (1 or 2). Default is 2.

Returns:

One or two offspring vectors.

Return type:

Union[ndarray, Tuple[ndarray, ndarray]]

Raises:

• ValueError – If parent vectors have different lengths or num_children

• is invalid. –

evolib.operators.crossover.crossover_intermediate(parent1, parent2, blend_range=0.25)

Perform intermediate crossover with extended alpha range on two parent vectors.

Each offspring gene is calculated using a random alpha ∈ [-d, 1 + d], creating solutions inside and outside the segment between parents.

Parameters:

• parent1 (ndarray) – First parent vector.

• parent2 (ndarray) – Second parent vector.

• d (float) – Extension factor for sampling interval beyond [0, 1].

• 0.25. (Default is)

Returns:

Two offspring vectors.

Return type:

Tuple[ndarray, ndarray]

Raises:

ValueError – If parent vectors have different lengths.

evolib.operators.crossover.crossover_offspring(pop, offspring)

Perform crossover for each pair of offspring individuals.

Delegates the actual crossover logic to the individuals’ parameter representations (ParaBase subclasses). Works for both single-module (e.g. Vector, EvoNet) and multi-module (ParaComposite) individuals.

Return type:

None

Notes

• Offspring are assumed to be copied before this call.

• Individuals are paired (0,1), (2,3), …

• This method does not return; offspring are modified in place.

evolib.operators.crossover.crossover_simulated_binary(parent1, parent2, eta=20)

Perform Simulated Binary Crossover (SBX) on two parent vectors.

SBX creates offspring that simulate the effect of single-point binary crossover in real-valued search spaces, controlled by a distribution index η.

Parameters:

• parent1 (ndarray) – First parent vector.

• parent2 (ndarray) – Second parent vector.

• eta (float) – Distribution index (controls spread; higher = closer to parents).

Returns:

Two offspring vectors.

Return type:

Tuple[ndarray, ndarray]

Raises:

ValueError – If parent vectors have different lengths.