Generation#

class empulse.optimizers.Generation(population_size=None, crossover_rate=0.8, mutation_rate=0.1, elitism=0.05, verbose=False, logging_fn=<built-in function print>, random_state=None, n_jobs=1)[source]#

A single generation of a Real-coded Genetic Algorithm (RGA).

optimize always maximizes the objective function passed to it. Read more in the User Guide.

Parameters:

population_sizeint or None, default=None: Number of individuals in the population. If None, population size is set to 10 * n_features.
crossover_ratefloat, default=0.8: Probability of crossover. Must be in [0, 1].
mutation_ratefloat, default=0.1: Probability of mutation. Must be in [0, 1].
elitismfloat, default=0.05: Fraction of the population that transferred to the next generation without change. Must be in [0, 1].
verbosebool, default=False: If True, print status messages.
logging_fncallable, default=print: Function to use for logging.
random_stateint, RandomState or None, default=None: Random seed. Accepts an int, a numpy.random.RandomState instance, or None (uses the global NumPy random state).
n_jobsint or None, default=1: Number of jobs to run in parallel. If -1, use all available processors. If None, use 1 processor.

Attributes:

namestr: Name of the optimizer.
directionDirection: Optimization direction, always Direction.MAXIMIZE. Callers handing this class a loss (which is minimized by convention) must negate it first.
populationndarray, shape (population_size, n_dim): Current population.
population_sizeint or None: The population_size constructor argument, unchanged by optimize. None means the actual size used each run is resolved from 10 * n_features and kept internally.
crossover_ratefloat: Probability of crossover.
mutation_ratefloat: Probability of mutation.
elitismint: The number of individuals of the population that are transferred to the next generation without change. Set to 0 until optimize resolves it from elitism_fraction once the population size is known.
elitism_fractionfloat: The elitism constructor argument, stored under its own name since elitism itself is repurposed as the resolved individual count above.
verbosebool: If True, print status messages.
logging_fncallable: Function to use for logging.
rngRandomState: Random state object.
n_jobsint: Number of jobs to run in parallel. If -1, use all available processors. If None, use 1 processor.
fx_bestlist: List of best fitness values.
fitnessndarray, shape (population_size,): Fitness values of the current population.
resultOptimizeResult: Result of the optimization.
lower_boundsndarray, shape (n_dim,): Lower bounds of the search space.
upper_boundsndarray, shape (n_dim,): Upper bounds of the search space.
delta_boundsndarray, shape (n_dim,): Difference between upper and lower bounds.
n_dimint: Number of dimensions.
_n_mating_pairsint: Number of mating pairs.
elite_poollist: List of elite individuals.

optimize(objective, bounds)[source]#

Optimize the objective function.

Parameters:

objectiveCallable: Objective function to optimize. Should be of signature objective(weights) -> float.
boundslist[tuple[float, float]]: List of tuples of lower and upper bounds for each weight.

Yields:

selfGeneration: Current instance of the optimizer.

Notes

This is an infinite generator. The caller is responsible for stopping iteration (e.g. via break or itertools.islice). Calling optimize on the same instance a second time resets all accumulated state (fx_best, result, elite_pool, etc.).