Source code for mealpy.swarm_based.ALO

# !/usr/bin/env python
# Created by "Thieu" at 12:01, 17/03/2020 ----------%
#       Email: nguyenthieu2102@gmail.com            %
#       Github: https://github.com/thieu1995        %
# --------------------------------------------------%

import numpy as np
from copy import deepcopy
from mealpy.optimizer import Optimizer


[docs]class OriginalALO(Optimizer): """ The original version of: Ant Lion Optimizer (ALO) Links: 1. https://www.mathworks.com/matlabcentral/fileexchange/49920-ant-lion-optimizer-alo 2. https://dx.doi.org/10.1016/j.advengsoft.2015.01.010 Examples ~~~~~~~~ >>> import numpy as np >>> from mealpy.swarm_based.ALO import OriginalALO >>> >>> def fitness_function(solution): >>> return np.sum(solution**2) >>> >>> problem_dict1 = { >>> "fit_func": fitness_function, >>> "lb": [-10, -15, -4, -2, -8], >>> "ub": [10, 15, 12, 8, 20], >>> "minmax": "min", >>> "verbose": True, >>> } >>> >>> epoch = 1000 >>> pop_size = 50 >>> model = OriginalALO(problem_dict1, epoch, pop_size) >>> best_position, best_fitness = model.solve() >>> print(f"Solution: {best_position}, Fitness: {best_fitness}") References ~~~~~~~~~~ [1] Mirjalili, S., 2015. The ant lion optimizer. Advances in engineering software, 83, pp.80-98. """ def __init__(self, problem, epoch=10000, pop_size=100, **kwargs): """ Args: problem (dict): The problem dictionary epoch (int): maximum number of iterations, default = 10000 pop_size (int): number of population size, default = 100 """ super().__init__(problem, kwargs) self.epoch = epoch self.pop_size = pop_size self.nfe_per_epoch = self.pop_size self.sort_flag = True def _random_walk_antlion(self, solution, current_epoch): I = 1 # I is the ratio in Equations (2.10) and (2.11) if current_epoch > self.epoch / 10: I = 1 + 100 * (current_epoch / self.epoch) if current_epoch > self.epoch / 2: I = 1 + 1000 * (current_epoch / self.epoch) if current_epoch > self.epoch * (3 / 4): I = 1 + 10000 * (current_epoch / self.epoch) if current_epoch > self.epoch * 0.9: I = 1 + 100000 * (current_epoch / self.epoch) if current_epoch > self.epoch * 0.95: I = 1 + 1000000 * (current_epoch / self.epoch) # Decrease boundaries to converge towards antlion lb = self.problem.lb / I # Equation (2.10) in the paper ub = self.problem.ub / I # Equation (2.10) in the paper # Move the interval of [lb ub] around the antlion [lb+anlion ub+antlion] if np.random.rand() < 0.5: lb = lb + solution # Equation(2.8) in the paper else: lb = -lb + solution if np.random.rand() < 0.5: ub = ub + solution # Equation(2.9) in the paper else: ub = -ub + solution # This function creates n random walks and normalize according to lb and ub vectors, temp = [] for k in range(0, self.problem.n_dims): X = np.cumsum(2 * (np.random.rand(self.epoch, 1) > 0.5) - 1) a = np.min(X) b = np.max(X) c = lb[k] # [a b] - -->[c d] d = ub[k] X_norm = ((X - a) * (d - c)) / (b - a) + c # Equation(2.7) in the paper temp.append(X_norm) return np.array(temp)
[docs] def evolve(self, epoch): """ The main operations (equations) of algorithm. Inherit from Optimizer class Args: epoch (int): The current iteration """ list_fitness = np.array([item[self.ID_TAR][self.ID_FIT] for item in self.pop]) # This for loop simulate random walks pop_new = [] for idx in range(0, self.pop_size): # Select ant lions based on their fitness (the better anlion the higher chance of catching ant) rolette_index = self.get_index_roulette_wheel_selection(list_fitness) # RA is the random walk around the selected antlion by rolette wheel RA = self._random_walk_antlion(self.pop[rolette_index][self.ID_POS], epoch) # RE is the random walk around the elite (best antlion so far) RE = self._random_walk_antlion(self.g_best[self.ID_POS], epoch) temp = (RA[:, epoch] + RE[:, epoch]) / 2 # Equation(2.13) in the paper # Bound checking (bring back the antlions of ants inside search space if they go beyonds the boundaries pos_new = self.amend_position(temp) pop_new.append([pos_new, None]) pop_new = self.update_fitness_population(pop_new) # Update antlion positions and fitnesses based of the ants (if an ant becomes fitter than an antlion we # assume it was caught by the antlion and the antlion update goes to its position to build the trap) self.pop = self.get_sorted_strim_population(self.pop + pop_new, self.pop_size) # Keep the elite in the population self.pop[-1] = deepcopy(self.g_best)
[docs]class BaseALO(OriginalALO): """ My changed version of: Ant Lion Optimizer (ALO) Notes ~~~~~ + Use matrix for better performance. + Change the flow of updating new position makes it better then original one Examples ~~~~~~~~ >>> import numpy as np >>> from mealpy.swarm_based.ALO import BaseALO >>> >>> def fitness_function(solution): >>> return np.sum(solution**2) >>> >>> problem_dict1 = { >>> "fit_func": fitness_function, >>> "lb": [-10, -15, -4, -2, -8], >>> "ub": [10, 15, 12, 8, 20], >>> "minmax": "min", >>> "verbose": True, >>> } >>> >>> epoch = 1000 >>> pop_size = 50 >>> model = BaseALO(problem_dict1, epoch, pop_size) >>> best_position, best_fitness = model.solve() >>> print(f"Solution: {best_position}, Fitness: {best_fitness}") """ def __init__(self, problem, epoch=10000, pop_size=100, **kwargs): """ Args: problem (dict): The problem dictionary epoch (int): maximum number of iterations, default = 10000 pop_size (int): number of population size, default = 100 """ super().__init__(problem, epoch, pop_size, **kwargs) self.nfe_per_epoch = self.pop_size self.sort_flag = True def _random_walk_antlion(self, solution, current_epoch): I = 1 # I is the ratio in Equations (2.10) and (2.11) if current_epoch > self.epoch / 10: I = 1 + 100 * (current_epoch / self.epoch) if current_epoch > self.epoch / 2: I = 1 + 1000 * (current_epoch / self.epoch) if current_epoch > self.epoch * (3 / 4): I = 1 + 10000 * (current_epoch / self.epoch) if current_epoch > self.epoch * 0.9: I = 1 + 100000 * (current_epoch / self.epoch) if current_epoch > self.epoch * 0.95: I = 1 + 1000000 * (current_epoch / self.epoch) # Dicrease boundaries to converge towards antlion lb = self.problem.lb / I # Equation (2.10) in the paper ub = self.problem.ub / I # Equation (2.10) in the paper # Move the interval of [lb ub] around the antlion [lb+anlion ub+antlion]. Eq 2.8, 2.9 lb = lb + solution if np.random.rand() < 0.5 else -lb + solution ub = ub + solution if np.random.rand() < 0.5 else -ub + solution # This function creates n random walks and normalize according to lb and ub vectors, ## Using matrix and vector for better performance X = np.array([np.cumsum(2 * (np.random.rand(self.epoch, 1) > 0.5) - 1) for _ in range(0, self.problem.n_dims)]) a = np.min(X, axis=1) b = np.max(X, axis=1) temp1 = np.reshape((ub - lb) / (b - a), (self.problem.n_dims, 1)) temp0 = X - np.reshape(a, (self.problem.n_dims, 1)) X_norm = temp0 * temp1 + np.reshape(lb, (self.problem.n_dims, 1)) return X_norm