Evolution

Package related to evolutionary agorithms.

Classes summary

`Evolution.Population.Population`(pb)	Class for the population of an evolutionary algorithm.
`Evolution.Mutation.Mutation`(prob)	Abstract class for mutation operators.
`Evolution.Polynomial.Polynomial`(prob, eta)	Class for polynomial mutation.
`Evolution.Crossover.Crossover`(prob)	Abstract class for crossover operators.
`Evolution.SBX.SBX`(prob, eta)	Class for SBX crossover.
`Evolution.Intermediate.Intermediate`([prob])	Class for intermediate crossover (weighted average of decision variables).
`Evolution.Two_Points.Two_Points`([prob])	Class for 2-point crossover.
`Evolution.Selection.Selection`()	Abstract class for selection operators.
`Evolution.Tournament.Tournament`(size)	Class for Tournament selection.
`Evolution.Tournament_Position.Tournament_Position`(size)	Class for Tournament_Position selection.
`Evolution.Replacement.Replacement`()	Abstract class for replacement operators.
`Evolution.Elitist.Elitist`()	Class for Elitist replacement.
`Evolution.Custom_Elitism.Custom_Elitism`(ec)	Class for custom elitist replacement.
`Evolution.Reference_Vector_Set.Reference_Vector_Set`(H, pb)	Class for the set of reference vectors of RVEA.

Population

class Evolution.Population.Population(pb)

Bases: object

Class for the population of an evolutionary algorithm.

Parameters:

pb (Problem) – problem
dvec (np.ndarray) – decision vectors of the individuals
obj_vals (np.ndarray) – objective value associated with each individual
fitness_modes (np.ndarray) – evaluation mode associated with each individual: True for real evaluation and False for prediction (surrogate evaluation)

__init__ method’s input

Parameters:: pb (Problem) – problem

print_shapes(): Prints the shapes of the arrays dvec, obj_vals and fitness_modes forming the population.

check_integrity()

Checks arrays’ shapes are consistent.

Returns:: True for arrays’ consistency and False otherwise
Return type:: bool

append(pop)

Appends individuals to the current population.

Parameters:: pop (Population) – individuals to be appended

sort(): Sorts the population according to ascending individuals’ objective value (single-objective) or non-dominated and crowded distance sorting (multi-objective).

split_in_batches(n_batch)

Splits the population in batches.

Parameters:: n_batch (positive int, not zero) – number of batches
Returns:: list of batches
Return type:: list(Population)

update_best_sim(f_best_profile, f_hypervolume=None)

Updates the best individual (single-objective) or the best non-dominated front (multi-objective) and logs.

For mono-objective: The best evaluated decision vector (minimisation assumed) is saved in Global_Var.dvec_min and its associated objective value is saved in Global_Var.obj_val_min. The best evaluated decision vector is printed to a file along with its associated objective value.

For multi-objective: If the hypervolume has improved, the evaluated decision vectors composing the best non-dominated front are printed to a file along with their respective objective value. The best hyper-volume is always printed to a file.

Parameters:

f_best_profile (str) – filename for logging
f_hypervolume (str) – filename for logging hypervolume

save_to_csv_file(f_pop_archive)

Prints the population to a CSV file.

The CSV file is organized as follows: First row: number of decision variables, number of objectives, number of fitness modes Second row: lower bounds of the decision variables Thrid row: upper bounds of the decision variables Remaining rows (one per individual): decision variables, objective values, fitness mode

Parameters:: f_pop_archive (str) – filename of the CSV file.

load_from_csv_file(f_pop_archive)

Loads the population from a CSV file.

The CSV file has to be organized as follows: First row: number of decision variables, number of objectives, number of fitness modes Second row: lower bounds of the decision variables Third row: upper bounds of the decision variables Remaining rows (one per individual): decision variables, objective values, fitness mode

Parameters:: f_pop_archive (str) – filename of the CSV file

save_sim_archive(f_sim_archive)

Prints the real-evaluated individuals to a CSV file.

The CSV file is organized as follows: One per individual: decision variables, objective values, fitness mode.

Parameters:: f_sim_archive (str) – filename of the CSV file.

save_to_pickle_file(f_pop_archive)

Saves the population to a pickle file.

Parameters:: f_pop_archive (str) – filename of the pickle file

load_from_pickle_file(f_pop_archive)

Loads a population from a pickle file.

Parameters:: f_pop_archive (str) – filename of the pickle file

Mutation

Mutation (abstract)

class Evolution.Mutation.Mutation(prob)

Bases: ABC

Abstract class for mutation operators.

Parameters:: prob (float in [0,1]) – probability of mutation

Polynomial

class Evolution.Polynomial.Polynomial(prob, eta)

Bases: Mutation

Class for polynomial mutation.

Parameters:

prob (float in [0,1]) – probability of mutation
eta (positive int, not zero) – distribution index

perform_mutation(pop)

Mutates the individuals of a population.

Parameters:: pop (Population) – population to mutate
Returns:: the mutated population
Return type:: Population

Crossover

Crossover (abstract)

class Evolution.Crossover.Crossover(prob)

Bases: ABC

Abstract class for crossover operators.

Parameters:: prob (float in [0,1]) – probability of crossover

SBX

class Evolution.SBX.SBX(prob, eta)

Bases: Crossover

Class for SBX crossover.

Parameters:

prob (float in [0,1]) – probability of crossover
eta (positive int, not zero) – distribution index

perform_crossover(pop)

Applies crossover to the individuals of a population.

Parameters:: pop (Population) – population to mutate
Returns:: the mutated population
Return type:: Population

Intermediate

class Evolution.Intermediate.Intermediate(prob=1.0)

Bases: Crossover

Class for intermediate crossover (weighted average of decision variables).

Parameters:: prob (float in [0,1]) – probability of crossover

perform_crossover(pop)

Applies crossover to the individuals of a population.

Parameters:: pop (Population) – population to mutate
Returns:: the crossed population
Return type:: Population

Two_Points

class Evolution.Two_Points.Two_Points(prob=1.0)

Bases: Crossover

Class for 2-point crossover.

Parameters:: prob (float in [0,1]) – probability of crossover

perform_crossover(pop)

Applies crossover to the individuals of a population.

Parameters:: pop (Population) – population of parents
Returns:: the crossed population
Return type:: Population

Selection

Selection (abstract)

class Evolution.Selection.Selection

Bases: ABC

Abstract class for selection operators.

abstract perform_selection(pop, n_par)

Selects individuals from a population.

Parameters:

pop (Population) – population to select from
n_par (positive int, not zero) – number of individuals to select

Returns:

the selected individuals

Return type:

Population

Tournament

class Evolution.Tournament.Tournament(size)

Bases: Selection

Class for Tournament selection.

Parameters:: size (positive int, not zero) – tournament size

perform_selection(pop, n_par)

Selects individuals from a population.

Parameters:

pop (Population) – population to select from
n_par (positive int, not zero) – number of individuals to select

Returns:

the selected individuals

Return type:

Population

Tournament_Position

class Evolution.Tournament_Position.Tournament_Position(size)

Bases: Selection

Class for Tournament_Position selection.

Candidates are ordered within the population according to their promise (decreasing order).

Parameters:: size (positive int, not zero) – tournament size

perform_selection(pop, n_par)

Selects individuals from a population.

Parameters:

pop (Population) – population to select from
n_par (positive int, not zero) – number of individuals to select

Returns:

the selected individuals

Return type:

Population

Replacement

Replacement (abstract)

class Evolution.Replacement.Replacement

Bases: ABC

Abstract class for replacement operators.

Elitist

class Evolution.Elitist.Elitist

Bases: Replacement

Class for Elitist replacement.

Minimization is assumed.

perform_replacement(pop, children)

Keeps the best individuals of from two populations.

Parameters:

pop (Population) – first population, will store the best individuals
children (Population) – second population

Custom_Elitism

class Evolution.Custom_Elitism.Custom_Elitism(ec)

Bases: Replacement

Class for custom elitist replacement.

Parameters:: ec (Surrogate) – the criterion defining elitism

perform_replacement(pop, children)

Keeps the best individuals from two populations.

Parameters:

pop (Population) – first population, will store the best individuals
children (Population) – second population

Reference Vector Set

class Evolution.Reference_Vector_Set.Reference_Vector_Set(H, pb)

Bases: object

Class for the set of reference vectors of RVEA.

Parameters:

pb (Problem) – problem
rv (np.ndarray) – set of reference vectors

__init__ method’s input

Parameters:

H (int) – simplex lattice parameter
pb (Problem) – problem

reference_vector_guided_replacement(pop, search_progress, budget)

Performs replacement based on the set of reference vectors.

Parameters:

pop (Population) – population to perform replacement on
search_progress (int or float) – current generation index (or elapsed time)
budget (int or float) – number of generations (or time) allocated for the search

Returns:

the new population

Return type:

Population

reference_vector_update(pop)

Updates the set of reference vectors.

Parameters:: pop (Population) – current population
Returns:: the updated set of reference vectors
Return type:: Reference_Vector_Set

reference_vector_regeneration(pop)

Regenerates the set of reference vectors (for RVEA*).

Parameters:: pop (Population) – current population

save_to_csv_file(f_rv_archive)

Prints the set of reference vectors to a CSV file.

The CSV file is organized as follows: First row: number of reference vectors and number of objectives Remaining rows: the reference vectors

Parameters:: f_rv_archive (str) – filename of the CSV file.

plot(): Plot the 3D reference vectors.