Source code for mealpy.human_based.HBO

#!/usr/bin/env python
# Created by "Thieu" at 00:47, 16/10/2022 ----------%                                                                               
#       Email: nguyenthieu2102@gmail.com            %                                                    
#       Github: https://github.com/thieu1995        %                         
# --------------------------------------------------%

import numpy as np
from mealpy.optimizer import Optimizer


[docs]class OriginalHBO(Optimizer): """ The original version of: Heap-based optimizer (HBO) Links: 1. https://www.sciencedirect.com/science/article/abs/pii/S0957417420305261#! 2. https://github.com/qamar-askari/HBO/blob/master/HBO.m Hyper-parameters should fine-tune in approximate range to get faster convergence toward the global optimum: + degree (int): [2, 4], the degree level in Corporate Rank Hierarchy (CRH), default=2 Examples ~~~~~~~~ >>> import numpy as np >>> from mealpy import FloatVar, HBO >>> >>> def objective_function(solution): >>> return np.sum(solution**2) >>> >>> problem_dict = { >>> "bounds": FloatVar(n_vars=30, lb=(-10.,) * 30, ub=(10.,) * 30, name="delta"), >>> "minmax": "min", >>> "obj_func": objective_function >>> } >>> >>> model = HBO.OriginalHBO(epoch=1000, pop_size=50, degree = 3) >>> g_best = model.solve(problem_dict) >>> print(f"Solution: {g_best.solution}, Fitness: {g_best.target.fitness}") >>> print(f"Solution: {model.g_best.solution}, Fitness: {model.g_best.target.fitness}") References ~~~~~~~~~~ [1] Askari, Q., Saeed, M., & Younas, I. (2020). Heap-based optimizer inspired by corporate rank hierarchy for global optimization. Expert Systems with Applications, 161, 113702. """ def __init__(self, epoch: int = 10000, pop_size: int = 100, degree: int = 2, **kwargs: object) -> None: """ Args: epoch (int): maximum number of iterations, default = 10000 pop_size (int): number of population size, default = 100 degree (int): the degree level in Corporate Rank Hierarchy (CRH), default=2 """ super().__init__(**kwargs) self.epoch = self.validator.check_int("epoch", epoch, [1, 100000]) self.pop_size = self.validator.check_int("pop_size", pop_size, [5, 10000]) self.degree = self.validator.check_int("degree", degree, [2, 10]) self.set_parameters(["epoch", "pop_size", "degree"]) self.is_parallelizable = False self.sort_flag = False
[docs] def initialize_variables(self): self.cycles = np.floor(self.epoch / 25) self.it_per_cycle = self.epoch / self.cycles self.qtr_cycle = self.it_per_cycle / 4
[docs] def colleagues_limits_generator__(self, pop_size, degree=3): friend_limits = np.zeros((pop_size, 2)) for c in range(pop_size - 1, -1, -1): hi = int(np.ceil((np.log10(c * degree - c + 1) / np.log10(degree)))) - 1 lowerLim = ((degree * degree ** (hi - 1) - 1) / (degree - 1) + 1) upperLim = (degree * degree ** hi - 1) / (degree - 1) friend_limits[c, 0] = lowerLim if lowerLim <= pop_size else pop_size friend_limits[c, 1] = upperLim if upperLim <= pop_size else pop_size return friend_limits.astype(int)
[docs] def heapifying__(self, pop, degree=3): pop_size = len(pop) heap = [] for c in range(pop_size): heap.append([pop[c].target, c]) # Heapifying t = c while t > 0: parent_id = int(np.floor((t + 1)/degree) - 1) if self.compare_target(pop[parent_id].target, pop[t].target, self.problem.minmax): break else: heap[t], heap[parent_id] = heap[parent_id], heap[t] t = parent_id return heap
[docs] def before_main_loop(self): self.heap = self.heapifying__(self.pop, self.degree) self.friend_limits = self.colleagues_limits_generator__(self.pop_size, self.degree)
[docs] def evolve(self, epoch): """ The main operations (equations) of algorithm. Inherit from Optimizer class Args: epoch (int): The current iteration """ gama = (np.mod(epoch, self.it_per_cycle) +1) / self.qtr_cycle gama = np.abs(2 - gama) p1 = 1. - epoch / self.epoch p2 = p1 + (1 - p1) / 2 for c in range(self.pop_size-1, 0, -1): if c == 0: # Dealing with root continue else: parent_id = int(np.floor((c+1)/self.degree) - 1) cur_agent = self.pop[self.heap[c][1]].copy() #Sol to be updated par_agent = self.pop[self.heap[parent_id][1]] #Sol to be updated with reference to # Sol to be updated with reference to if self.friend_limits[c, 0] < self.friend_limits[c, 1]+1: friend_idx = self.friend_limits[c, 0] else: friend_idx = self.generator.choice(list(set(range(self.friend_limits[c, 0], self.friend_limits[c, 1])) - {c})) fri_agent = self.pop[self.heap[friend_idx][1]] #Position Updating rr = self.generator.random(self.problem.n_dims) rn = (2 * self.generator.random(self.problem.n_dims) - 1) for jdx in range(self.problem.n_dims): if rr[jdx] < p1: continue elif rr[jdx] < p2: cur_agent.solution[jdx] = par_agent.solution[jdx] + rn[jdx] * gama * np.abs(par_agent.solution[jdx] - cur_agent.solution[jdx]) else: if self.compare_target(self.heap[friend_idx][0], self.heap[c][0], self.problem.minmax): cur_agent.solution[jdx] = fri_agent.solution[jdx] + rn[jdx] * gama * np.abs(fri_agent.solution[jdx] - cur_agent.solution[jdx]) else: cur_agent.solution[jdx] += rn[jdx] * gama * np.abs(fri_agent.solution[jdx] - cur_agent.solution[jdx]) pos_new = self.correct_solution(cur_agent.solution) cur_agent = self.generate_agent(pos_new) if self.compare_target(cur_agent.target, self.heap[c][0], self.problem.minmax): self.pop[self.heap[c][1]] = cur_agent self.heap[c][0] = cur_agent.target.copy() # Heapifying t = c while t > 1: parent_id = int((t + 1) / self.degree) if self.compare_target(self.heap[parent_id][0], self.heap[t][0], self.problem.minmax): break else: self.heap[t], self.heap[parent_id] = self.heap[parent_id], self.heap[t] t = parent_id