def __init__(self,
filename,
population_size=100,
mutation_probability=0.1,
pool_probability=0.2,
timer_limit=20000):
# Matrices
self.weights = get_weights_matrix(filename, static=False)
#self.weights = self.weights[:50, :50, :50]
# constants
self.N_CITIES = np.shape(self.weights)[1]
self.TIMER_LIMIT = timer_limit
self.MUTATION_PROBABILITY = mutation_probability
self.POOL_PROBABILITY = pool_probability
self.POP_SIZE = population_size
self.SENSIBILITY = 0.1
# population
self.population = init_population(self.POP_SIZE, self.N_CITIES)
self.population_cost = None # it is a numpy array.
self.compute_population_cost()
# gene_pool
self.K = 6 # cardinality of the gene pool
self.gene_pool = []
self.init_gene_pool()
# solutions:
self.best_cost = float('Inf')
self.best_solution = None
def __init__(self,
filename,
card_max=None,
is_static=True,
is_fengyun=False):
if is_fengyun:
self.weights, self.N_CITIES = get_fengyun_weights(filename)
else:
self.weights, self.N_CITIES = get_weights_matrix(
filename, card_max=card_max, is_static=is_static)
self.is_static = is_static
self.best_solution = None
self.best_cost = float('Inf')
def __init__(self,
filename,
n_ants=10,
alpha=0.1,
beta=6,
rho=0.1,
q0=0.98,
max_iteration=100):
# Matrices
self.weights = get_weights_matrix(filename, static=True)
#self.weights = self.weights[:20, :20]
# Constants
self.BETA = beta
self.N_CITIES = np.shape(self.weights)[1]
self.N_ANTS = n_ants
self.MAX_ITERATION = max_iteration
self.TAO_0 = 0
self.RHO = rho
self.ALPHA = alpha
self.Q0 = q0
# Solutions
self.solutions = []
self.new_solutions = None
# global solution
self.global_best_solution = None
self.global_best_cost = float('Inf')
# indexes
self.iteration = 0
# self.start_idx = 0
# pheromone
self.pheromone_matrix = None
self.init_pheromone_matrix()
def __init__(self,
filename,
weights=None,
population=None,
is_static=True,
population_size=100,
mutation_probability=0.05,
timer_limit=20000):
self.is_static = is_static
# Matrices
if weights is None:
self.weights = get_weights_matrix(filename, static=self.is_static)
else:
self.weights = weights
#self.weights = self.weights[:25, :25]
# population
if population is None:
self.population = None
self.POP_SIZE = population_size
else:
self.population = population
self.POP_SIZE = len(self.population)
self.population_cost = None
self.compute_population_cost()
# constants
self.N_CITIES = np.shape(self.weights)[1]
self.TIMER_LIMIT = timer_limit
self.MUTATION_PROBABILITY = mutation_probability # mutation probability
# solutions:
self.best_cost = float('Inf')
self.best_solution = None
self.nn_solution = []
self.nn_cost = float('Inf')
def __init__(self,
filename,
card_max=None,
n_ants=5,
alpha=1,
beta=6,
rho=0.1,
q0=0.9,
max_generations=1000,
type="AS",
is_fengyun=False):
"""Constructor of the class Dynamic Ant.
Parameters:
filename : string that contains the path of the weights file.
card_max : (optional - default None) integer that represent the maximum number of cities to use.
n_ants : (optional - default 5) integer. The number of the ants. .
alpha : (optional - default 1) pheromone trail influence. The greater the more important is the pheromone
trail.
beta : (optional - dafault 6) heuristic information influence. The greater the more important is the
heuristic.
rho : (optional - default 0.1) pheromone evaporation rate.
q0 : (optional - default 0.9) probability threshold of choosing the next city with the probabilistic rule.
max_generations : (optional - default 1000) maximum number of the generations.
type : (optional - default AS) Specify which type of approach keeping during pheromone deposit.
is_fengyun : (optional - default False) If set to true change the ACO behaviour in order to manage its huge dataset.
"""
# Weights Matrix
if is_fengyun:
self.weights, self.N_CITIES = get_fengyun_weights(filename)
else:
self.weights, self.N_CITIES = get_weights_matrix(filename,
card_max=card_max,
is_static=False)
# Constants
self.BETA = beta
self.N_ANTS = n_ants
self.MAX_GENERATIONS = max_generations
self.RHO = rho # pheromone evaporation speed.
self.ALPHA = alpha
self.Q0 = q0 # to be global constant
self.TYPE = type
self.is_fengyun = is_fengyun
# Precompute the heuristic information based on the weights.
if not self.is_fengyun:
self.heuristic_info = np.power(
np.divide(1, self.weights, where=self.weights != 0), self.BETA)
# Time iterator for the dynamic application.
self.t = 0 # current time
self.t0 = 0 # initial time
self.T = np.shape(self.weights)[0] - 1 # final time (index)
# Solutions
self.solutions = []
self.new_solutions = None # new solutions.
self.new_costs = None # costs of the new solutions.
self.nn_solution = []
self.nn_cost = float('Inf')
# global solution
self.global_best_solution = None
self.global_best_cost = float('Inf')
# indexes
self.iteration = 0
# pheromone
self.TAO_0 = None # minimum pheromone level.
self.pheromone_matrix = None
self.init_pheromone_matrix()
# list of length k of sets of available cities
self.available = None
# aio local search
self.IO_MAX_ITER = 10
self.p_bi = 0.5
self.p_gi = 0.5
# statistic variables
self.best_performance = np.zeros(self.MAX_GENERATIONS)
self.avg_performance = np.zeros(self.MAX_GENERATIONS)
self.worst_performance = np.zeros(self.MAX_GENERATIONS)
self.change_global_min = np.zeros(self.MAX_GENERATIONS)
self.global_p_bi = np.zeros(self.MAX_GENERATIONS)
self.global_p_gi = np.zeros(self.MAX_GENERATIONS)
self.global_n_bi_inversions = np.zeros(self.MAX_GENERATIONS)
self.global_n_gi_inversions = np.zeros(self.MAX_GENERATIONS)
self.relative_diversity = np.zeros(self.MAX_GENERATIONS)
self.avg_time_per_solution = np.zeros(self.MAX_GENERATIONS)
self.ls_number_of_improvements = np.zeros(
(self.N_ANTS, self.MAX_GENERATIONS))
self.ls_rate_of_improvements = np.zeros(
(self.N_ANTS, self.MAX_GENERATIONS))
def __init__(self,
filename,
card_max=None,
n_ants=5,
alpha=1,
beta=6,
q0=0.9,
pop_size=100,
max_generation=1000):
"""Constructor of the class Dynamic Ant for the Population Based Ant Colony Optimization.
Parameters:
filename : string that contains the path of the weights file.
card_max : (optional - default None) integer that represent the maximum number of cities to use.
n_ants : (optional - default 5) integer. The number of the ants. .
alpha : (optional - default 1) pheromone trail influence. The greater the more important is the pheromone
trail.
beta : (optional - dafault 1) heuristic information influence. The greater the more important is the
heuristic.
q0 : (optional - default 0.1) probability threshold of choosing the next city with the probabilistic rule.
max_generation : (optional - default 1000) maximum number of the generations.
"""
# Weights Matrix
self.weights = get_weights_matrix(filename)
if card_max is not None and card_max < np.shape(
self.weights)[1] and card_max > 0:
self.N_CITIES = card_max
self.weights = self.weights[:card_max, :card_max, :card_max]
else:
self.N_CITIES = np.shape(self.weights)[1]
# Constants
self.BETA = beta
self.N_ANTS = n_ants
self.MAX_GENERATION = max_generation
self.TAO_0 = 0 # minimum pheromone level. initialized later.
self.ALPHA = alpha
self.Q0 = q0 # to be global constant
self.MAX_POP = pop_size
# Precompute the heuristic information based on the weights.
self.heuristic_info = np.power(
np.divide(1, self.weights, where=self.weights != 0), self.BETA)
# Time iterator for the dynamic application.
self.t = 0 # current time
self.t0 = 0 # initial time
self.T = np.shape(self.weights)[0] - 1 # final time (index)
# Solutions
self.solutions = []
self.new_solutions = None
self.nn_solution = []
self.nn_cost = float('Inf')
# global solution
self.global_best_solution = None
self.global_best_cost = float('Inf')
# indexes
self.iteration = 0
# pheromone
self.TAO_0 = None # minimum pheromone level.
self.TAO_MAX = None # maximum pheromone level.
self.TAO_DELTA = None # delta step for the pheromone.
self.pheromone_matrix = None
self.init_pheromone_matrix()
# list of length k of sets of available cities
self.available = None
# population: managed with the age mechanism.
self.population = []
self.pop_iter = 0
self.pop_iter_age = 0
self.overflow = 0