#!/usr/bin/env python
# Created by "Thieu" at 20:22, 12/06/2020 ----------%
# Email: nguyenthieu2102@gmail.com %
# Github: https://github.com/thieu1995 %
# --------------------------------------------------%
import numpy as np
from mealpy.optimizer import Optimizer
[docs]class DevSMA(Optimizer):
"""
The developed version: Slime Mould Algorithm (SMA)
Notes:
+ Selected 2 unique and random solution to create new solution (not to create variable)
+ Check bound and compare old position with new position to get the best one
Hyper-parameters should fine-tune in approximate range to get faster convergence toward the global optimum:
+ p_t (float): (0, 1.0) -> better [0.01, 0.1], probability threshold (z in the paper)
Examples
~~~~~~~~
>>> import numpy as np
>>> from mealpy import FloatVar, SMA
>>>
>>> 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 = SMA.DevSMA(epoch=1000, pop_size=50, p_t = 0.03)
>>> 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}")
"""
def __init__(self, epoch: int = 10000, pop_size: int = 100, p_t: float = 0.03, **kwargs: object) -> None:
"""
Args:
epoch (int): maximum number of iterations, default = 10000
pop_size (int): number of population size, default = 100
p_t (float): probability threshold (z in the paper), default = 0.03
"""
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.p_t = self.validator.check_float("p_t", p_t, (0, 1.0))
self.set_parameters(["epoch", "pop_size", "p_t"])
self.sort_flag = True
[docs] def initialize_variables(self):
self.weights = np.zeros((self.pop_size, self.problem.n_dims))
[docs] def evolve(self, epoch):
"""
The main operations (equations) of algorithm. Inherit from Optimizer class
Args:
epoch (int): The current iteration
"""
# plus eps to avoid denominator zero
ss = self.g_best.target.fitness - self.pop[-1].target.fitness + self.EPSILON
# calculate the fitness weight of each slime mold
for idx in range(0, self.pop_size):
# Eq.(2.5)
if idx <= int(self.pop_size / 2):
self.weights[idx] = 1 + self.generator.uniform(0, 1, self.problem.n_dims) * \
np.log10((self.g_best.target.fitness - self.pop[idx].target.fitness) / ss + 1)
else:
self.weights[idx] = 1 - self.generator.uniform(0, 1, self.problem.n_dims) * \
np.log10((self.g_best.target.fitness - self.pop[idx].target.fitness) / ss + 1)
a = np.arctanh(-(epoch / self.epoch) + 1) # Eq.(2.4)
b = 1 - epoch / self.epoch
pop_new = []
for idx in range(0, self.pop_size):
# Update the Position of search agent
if self.generator.uniform() < self.p_t: # Eq.(2.7)
pos_new = self.problem.generate_solution()
else:
p = np.tanh(np.abs(self.pop[idx].target.fitness - self.g_best.target.fitness)) # Eq.(2.2)
vb = self.generator.uniform(-a, a, self.problem.n_dims) # Eq.(2.3)
vc = self.generator.uniform(-b, b, self.problem.n_dims)
# two positions randomly selected from population, apply for the whole problem size instead of 1 variable
id_a, id_b = self.generator.choice(list(set(range(0, self.pop_size)) - {idx}), 2, replace=False)
pos_1 = self.g_best.solution + vb * (self.weights[idx] * self.pop[id_a].solution - self.pop[id_b].solution)
pos_2 = vc * self.pop[idx].solution
condition = self.generator.random(self.problem.n_dims) < p
pos_new = np.where(condition, pos_1, pos_2)
# Check bound and re-calculate fitness after each individual move
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)
[docs]class OriginalSMA(DevSMA):
"""
The original version of: Slime Mould Algorithm (SMA)
Links:
1. https://doi.org/10.1016/j.future.2020.03.055
2. https://www.researchgate.net/publication/340431861_Slime_mould_algorithm_A_new_method_for_stochastic_optimization
Hyper-parameters should fine-tune in approximate range to get faster convergence toward the global optimum:
+ p_t (float): (0, 1.0) -> better [0.01, 0.1], probability threshold (z in the paper)
Examples
~~~~~~~~
>>> import numpy as np
>>> from mealpy import FloatVar, SMA
>>>
>>> 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 = SMA.OriginalSMA(epoch=1000, pop_size=50, p_t = 0.03)
>>> 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] Li, S., Chen, H., Wang, M., Heidari, A.A. and Mirjalili, S., 2020. Slime mould algorithm: A new method for
stochastic optimization. Future Generation Computer Systems, 111, pp.300-323.
"""
def __init__(self, epoch=10000, pop_size=100, p_t=0.03, **kwargs):
"""
Args:
epoch (int): maximum number of iterations, default = 1000
pop_size (int): number of population size, default = 100
p_t (float): probability threshold (z in the paper), default = 0.03
"""
super().__init__(epoch, pop_size, p_t, **kwargs)
[docs] def evolve(self, epoch):
"""
The main operations (equations) of algorithm. Inherit from Optimizer class
Args:
epoch (int): The current iteration
"""
# plus eps to avoid denominator zero
ss = self.g_best.target.fitness - self.pop[-1].target.fitness + self.EPSILON
# calculate the fitness weight of each slime mold
for idx in range(0, self.pop_size):
# Eq.(2.5)
if idx <= int(self.pop_size / 2):
self.weights[idx] = 1 + self.generator.uniform(0, 1, self.problem.n_dims) * \
np.log10((self.g_best.target.fitness - self.pop[idx].target.fitness) / ss + 1)
else:
self.weights[idx] = 1 - self.generator.uniform(0, 1, self.problem.n_dims) * \
np.log10((self.g_best.target.fitness - self.pop[idx].target.fitness) / ss + 1)
aa = np.arctanh(-(epoch / self.epoch) + 1) # Eq.(2.4)
bb = 1 - epoch / self.epoch
pop_new = []
for idx in range(0, self.pop_size):
# Update the Position of search agent
pos_new = self.pop[idx].solution.copy()
if self.generator.uniform() < self.p_t: # Eq.(2.7)
pos_new = self.problem.generate_solution()
else:
p = np.tanh(np.abs(self.pop[idx].target.fitness - self.g_best.target.fitness)) # Eq.(2.2)
vb = self.generator.uniform(-aa, aa, self.problem.n_dims) # Eq.(2.3)
vc = self.generator.uniform(-bb, bb, self.problem.n_dims)
for jdx in range(0, self.problem.n_dims):
# two positions randomly selected from population
id_a, id_b = self.generator.choice(list(set(range(0, self.pop_size)) - {idx}), 2, replace=False)
if self.generator.uniform() < p: # Eq.(2.1)
pos_new[jdx] = self.g_best.solution[jdx] + vb[jdx] * (self.weights[idx, jdx] * self.pop[id_a].solution[jdx] - self.pop[id_b].solution[jdx])
else:
pos_new[jdx] = vc[jdx] * pos_new[jdx]
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] = agent
if self.mode in self.AVAILABLE_MODES:
self.pop = self.update_target_for_population(pop_new)