modea.Selection module

This module contains a collection of Selection operators.

Selection accepts (mu + lambda) individuals and returns (mu) individuals that are chosen to be the best of this generation according to which selection module is chosen.

modea.Selection.best(population, new_population, param)

Given the population, return the (mu) best. Also performs some ‘housekeeping’ for the CMA-ES by collecting all genotypes and most recent mutation vectors and storing them in the param object.

Parameters:
  • population – List of FloatIndividual objects containing the previous generation
  • new_population – List of FloatIndividual objects containing the new generation
  • param – Parameters object for storing all parameters, options, etc.
Returns:

A slice of the sorted new_population list.
modea.Selection.bestGA(population, new_population, param)

Given the population, return the (mu) best

Parameters:
  • population – List of MixedIntIndividual objects containing the previous generation
  • new_population – List of MixedIntIndividual objects containing the new generation
  • param – Parameters object for storing all parameters, options, etc.
Returns:

A slice of the sorted new_population list.
modea.Selection.onePlusOneSelection(population, new_population, t, param)

(1+1)-selection (with success history)

Parameters:
  • population – List of FloatIndividual objects containing the previous generation
  • new_population – List of FloatIndividual objects containing the new generation
  • t – Timestamp of the current generation being evaluated
  • param – Parameters object for storing all parameters, options, etc.
Returns:

A slice of the sorted new_population list.
modea.Selection.pairwise(population, new_population, param)

Perform a selection on individuals in a population per pair, before letting best() make the final selection. Intended for use with a MirroredSampling sampler to prevent step-size bias.

Assumes that new_population contains pairs as [P1_a, P1_b, P2_a, P2_b, etc … ]

Parameters:
  • population – List of FloatIndividual objects containing the previous generation
  • new_population – List of FloatIndividual objects containing the new generation
  • param – Parameters object for storing all parameters, options, etc.
Returns:

A slice of the sorted new_population list.
modea.Selection.roulette(population, new_population, param, force_unique=False)

Given the population, return mu individuals, selected by roulette, using 1/fitness as probability

Parameters:
  • population – List of FloatIndividual objects containing the previous generation
  • new_population – List of FloatIndividual objects containing the new generation
  • param – Parameters object for storing all parameters, options, etc.
  • force_unique – Determine if an individual from the original population may be selected multiple times
Returns:

A slice of the sorted new_population list.