#!/usr/bin/env python
# Created by "Thieu" at 18:08, 19/04/2020 ----------%
# Email: nguyenthieu2102@gmail.com %
# Github: https://github.com/thieu1995 %
# --------------------------------------------------%
import numpy as np
from mealpy.optimizer import Optimizer
[docs]class OriginalCEM(Optimizer):
"""
The original version of: Cross-Entropy Method (CEM)
Links:
1. https://github.com/clever-algorithms/CleverAlgorithms
2. https://doi.org/10.1007/s10479-005-5724-z
Hyper-parameters should fine-tune in approximate range to get faster convergence toward the global optimum:
+ n_best (int): N selected solutions as a samples for next evolution
+ alpha (float): weight factor for means and stdevs (normal distribution)
Examples
~~~~~~~~
>>> import numpy as np
>>> from mealpy import FloatVar, CEM
>>>
>>> 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 = CEM.OriginalCEM(epoch=1000, pop_size=50, n_best = 20, alpha = 0.7)
>>> 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] De Boer, P.T., Kroese, D.P., Mannor, S. and Rubinstein, R.Y., 2005. A tutorial on the
cross-entropy method. Annals of operations research, 134(1), pp.19-67.
"""
def __init__(self, epoch: int = 10000, pop_size: int = 100, n_best: int = 20, alpha: float = 0.7, **kwargs: object) -> None:
"""
Args:
epoch (int): maximum number of iterations, default = 10000
pop_size (int): number of population size, default = 100
n_best (int): N selected solutions as a samples for next evolution
alpha (float): weight factor for means and stdevs (normal distribution)
"""
super().__init__(**kwargs)
self.epoch = self.validator.check_int("epoch", epoch, [1, 100000])
self.pop_size = self.validator.check_int("pop_size", pop_size, [10, 10000])
self.n_best = self.validator.check_int("n_best", n_best, [2, int(self.pop_size/2)])
self.alpha = self.validator.check_float("alpha", alpha, (0, 1.0))
self.set_parameters(["epoch", "pop_size", "n_best", "alpha"])
self.sort_flag = True
[docs] def initialize_variables(self):
self.means = self.generator.uniform(self.problem.lb, self.problem.ub)
self.stdevs = np.abs(self.problem.ub - self.problem.lb)
[docs] def evolve(self, epoch):
"""
The main operations (equations) of algorithm. Inherit from Optimizer class
Args:
epoch (int): The current iteration
"""
## Selected the best samples and update means and stdevs
pop_best = self.pop[:self.n_best]
pos_list = np.array([agent.solution for agent in pop_best])
means_new = np.mean(pos_list, axis=0)
means_new_repeat = np.repeat(means_new.reshape((1, -1)), self.n_best, axis=0)
stdevs_new = np.mean((pos_list - means_new_repeat) ** 2, axis=0)
self.means = self.alpha * self.means + (1.0 - self.alpha) * means_new
self.stdevs = np.abs(self.alpha * self.stdevs + (1.0 - self.alpha) * stdevs_new)
## Create new population for next generation
pop_new = []
for idx in range(0, self.pop_size):
pos_new = self.generator.normal(self.means, self.stdevs)
pos_new = self.correct_solution(pos_new)
agent = self.generate_empty_agent(pos_new)
pop_new.append(agent)
if self.mode not in self.AVAILABLE_MODES:
agent.target = self.get_target(pos_new)
self.pop[idx] =self.get_better_agent(agent, self.pop[idx], self.problem.minmax)
if self.mode in self.AVAILABLE_MODES:
pop_new = self.update_target_for_population(pop_new)
self.pop = self.greedy_selection_population(self.pop, pop_new, self.problem.minmax)