Source code for mealpy.human_based.TLO

# !/usr/bin/env python
# Created by "Thieu" at 10:14, 18/03/2020 ----------%
#       Email: nguyenthieu2102@gmail.com            %
#       Github: https://github.com/thieu1995        %
# --------------------------------------------------%

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


[docs]class BaseTLO(Optimizer): """ The original version of: Teaching Learning-based Optimization (TLO) Links: 1. https://doi.org/10.5267/j.ijiec.2012.03.007 Notes ~~~~~ + Removed the third loop to make it faster + This version taken the advantages of numpy np.array to faster handler operations + The global best solution is used Examples ~~~~~~~~ >>> import numpy as np >>> from mealpy.human_based.TLO import BaseTLO >>> >>> 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 = BaseTLO(problem_dict1, epoch, pop_size) >>> best_position, best_fitness = model.solve() >>> print(f"Solution: {best_position}, Fitness: {best_fitness}") References ~~~~~~~~~~ [1] Rao, R. and Patel, V., 2012. An elitist teaching-learning-based optimization algorithm for solving complex constrained optimization problems. international journal of industrial engineering computations, 3(4), pp.535-560. """ 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.nfe_per_epoch = 2 * pop_size self.sort_flag = False self.epoch = epoch self.pop_size = pop_size
[docs] def evolve(self, epoch): """ The main operations (equations) of algorithm. Inherit from Optimizer class Args: epoch (int): The current iteration """ pop_new = [] for idx in range(0, self.pop_size): ## Teaching Phrase TF = np.random.randint(1, 3) # 1 or 2 (never 3) list_pos = np.array([item[self.ID_POS] for item in self.pop]) DIFF_MEAN = np.random.rand(self.problem.n_dims) * (self.g_best[self.ID_POS] - TF * np.mean(list_pos, axis=0)) temp = self.pop[idx][self.ID_POS] + DIFF_MEAN pos_new = self.amend_position(temp) pop_new.append([pos_new, None]) pop_new = self.update_fitness_population(pop_new) pop_new = self.greedy_selection_population(self.pop, pop_new) pop_child = [] for idx in range(0, self.pop_size): ## Learning Phrase temp = deepcopy(pop_new[idx][self.ID_POS]).astype(float) id_partner = np.random.choice(np.setxor1d(np.array(range(self.pop_size)), np.array([idx]))) if self.compare_agent(pop_new[idx], pop_new[id_partner]): temp += np.random.rand(self.problem.n_dims) * (pop_new[idx][self.ID_POS] - pop_new[id_partner][self.ID_POS]) else: temp += np.random.rand(self.problem.n_dims) * (pop_new[id_partner][self.ID_POS] - pop_new[idx][self.ID_POS]) pos_new = self.amend_position(temp) pop_child.append([pos_new, None]) pop_child = self.update_fitness_population(pop_child) self.pop = self.greedy_selection_population(pop_new, pop_child)
[docs]class OriginalTLO(BaseTLO): """ The original version of: Teaching Learning-based Optimization (TLO) Links: 1. https://github.com/andaviaco/tblo Notes ~~~~~ + Removed the third loop to make it faster + This is slower version which inspired from link below Examples ~~~~~~~~ >>> import numpy as np >>> from mealpy.human_based.TLO import OriginalTLO >>> >>> 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 = OriginalTLO(problem_dict1, epoch, pop_size) >>> best_position, best_fitness = model.solve() >>> print(f"Solution: {best_position}, Fitness: {best_fitness}") References ~~~~~~~~~~ [1] Rao, R.V., Savsani, V.J. and Vakharia, D.P., 2011. Teaching–learning-based optimization: a novel method for constrained mechanical design optimization problems. Computer-aided design, 43(3), pp.303-315. """ 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 = 2 * pop_size self.sort_flag = False
[docs] def evolve(self, epoch): """ The main operations (equations) of algorithm. Inherit from Optimizer class Args: epoch (int): The current iteration """ for idx in range(0, self.pop_size): ## Teaching Phrase TF = np.random.randint(1, 3) # 1 or 2 (never 3) #### Remove third loop here list_pos = np.array([item[self.ID_POS] for item in self.pop]) pos_new = self.pop[idx][self.ID_POS] + np.random.uniform(0, 1, self.problem.n_dims) * \ (self.g_best[self.ID_POS] - TF * np.mean(list_pos, axis=0)) pos_new = self.amend_position(pos_new) fit_new = self.get_fitness_position(pos_new) if self.compare_agent([pos_new, fit_new], self.pop[idx]): self.pop[idx] = [pos_new, fit_new] ## Learning Phrase id_partner = np.random.choice(np.setxor1d(np.array(range(self.pop_size)), np.array([idx]))) #### Remove third loop here if self.compare_agent(self.pop[idx], self.pop[id_partner]): diff = self.pop[idx][self.ID_POS] - self.pop[id_partner][self.ID_POS] else: diff = self.pop[id_partner][self.ID_POS] - self.pop[idx][self.ID_POS] pos_new = self.pop[idx][self.ID_POS] + np.random.uniform(0, 1, self.problem.n_dims) * diff pos_new = self.amend_position(pos_new) fit_new = self.get_fitness_position(pos_new) if self.compare_agent([pos_new, fit_new], self.pop[idx]): self.pop[idx] = [pos_new, fit_new]
[docs]class ITLO(BaseTLO): """ The original version of: Improved Teaching-Learning-based Optimization (ITLO) Links: 1. https://doi.org/10.1016/j.scient.2012.12.005 Notes ~~~~~ + Removed the third loop to make it faster + Kinda similar to the paper, but the pseudo-code in the paper is not clear. Hyper-parameters should fine tuned in approximate range to get faster convergen toward the global optimum: + n_teachers (int): [3, 10], number of teachers in class, default=5 Examples ~~~~~~~~ >>> import numpy as np >>> from mealpy.human_based.TLO import ITLO >>> >>> 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 = ITLO(problem_dict1, epoch, pop_size) >>> best_position, best_fitness = model.solve() >>> print(f"Solution: {best_position}, Fitness: {best_fitness}") References ~~~~~~~~~~ [1] Rao, R.V. and Patel, V., 2013. An improved teaching-learning-based optimization algorithm for solving unconstrained optimization problems. Scientia Iranica, 20(3), pp.710-720. """ def __init__(self, problem, epoch=10000, pop_size=100, n_teachers=5, **kwargs): """ Args: problem (dict): The problem dictionary epoch (int): maximum number of iterations, default = 10000 pop_size (int): number of population size, default = 100 n_teachers (int): number of teachers in class """ super().__init__(problem, epoch, pop_size, **kwargs) self.nfe_per_epoch = 2 * pop_size self.n_teachers = n_teachers # Number of teams / group self.n_students = pop_size - n_teachers self.n_students_in_team = int(self.n_students / self.n_teachers) self.teachers, self.teams = None, None
[docs] def classify(self, pop): sorted_pop, best = self.get_global_best_solution(pop) teachers = sorted_pop[:self.n_teachers] sorted_pop = sorted_pop[self.n_teachers:] idx_list = np.random.permutation(range(0, self.n_students)) teams = [] for id_teacher in range(0, self.n_teachers): group = [] for idx in range(0, self.n_students_in_team): start_index = id_teacher * self.n_students_in_team + idx group.append(sorted_pop[idx_list[start_index]]) teams.append(group) return teachers, teams, best
[docs] def initialization(self): self.pop = self.create_population(self.pop_size) self.teachers, self.teams, self.g_best = self.classify(self.pop)
[docs] def evolve(self, epoch): """ The main operations (equations) of algorithm. Inherit from Optimizer class Args: epoch (int): The current iteration """ for id_teach, teacher in enumerate(self.teachers): team = self.teams[id_teach] list_pos = np.array([student[self.ID_POS] for student in self.teams[id_teach]]) # Step 7 mean_team = np.mean(list_pos, axis=0) pop_new = [] for id_stud, student in enumerate(team): if teacher[self.ID_TAR][self.ID_FIT] == 0: TF = 1 else: TF = student[self.ID_TAR][self.ID_FIT] / teacher[self.ID_TAR][self.ID_FIT] diff_mean = np.random.rand() * (teacher[self.ID_POS] - TF * mean_team) # Step 8 id2 = np.random.choice(list(set(range(0, self.n_teachers)) - {id_teach})) if self.compare_agent(teacher, team[id2]): pos_new = (student[self.ID_POS] + diff_mean) + np.random.rand() * (team[id2][self.ID_POS] - student[self.ID_POS]) else: pos_new = (student[self.ID_POS] + diff_mean) + np.random.rand() * (student[self.ID_POS] - team[id2][self.ID_POS]) pos_new = self.amend_position(pos_new) pop_new.append([pos_new, None]) pop_new = self.update_fitness_population(pop_new) self.teams[id_teach] = self.greedy_selection_population(team, pop_new) for id_teach, teacher in enumerate(self.teachers): ef = round(1 + np.random.rand()) team = self.teams[id_teach] pop_new = [] for id_stud, student in enumerate(team): id2 = np.random.choice(list(set(range(0, self.n_students_in_team)) - {id_stud})) if self.compare_agent(student, team[id2]): pos_new = student[self.ID_POS] + np.random.rand() * (student[self.ID_POS] - team[id2][self.ID_POS]) + \ np.random.rand() * (teacher[self.ID_POS] - ef * team[id2][self.ID_POS]) else: pos_new = student[self.ID_POS] + np.random.rand() * (team[id2][self.ID_POS] - student[self.ID_POS]) + \ np.random.rand() * (teacher[self.ID_POS] - ef * student[self.ID_POS]) pos_new = self.amend_position(pos_new) pop_new.append([pos_new, None]) pop_new = self.update_fitness_population(pop_new) self.teams[id_teach] = self.greedy_selection_population(team, pop_new) for id_teach, teacher in enumerate(self.teachers): team = self.teams[id_teach] + [teacher] team, local_best = self.get_global_best_solution(team) self.teachers[id_teach] = local_best self.teams[id_teach] = team[1:] self.pop = self.teachers + reduce(lambda x, y: x + y, self.teams)