# This file is part of NEORL.
# Copyright (c) 2021 Exelon Corporation and MIT Nuclear Science and Engineering
# NEORL is free software: you can redistribute it and/or modify
# it under the terms of the MIT LICENSE
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
# -*- coding: utf-8 -*-
#"""
#Created on Mon Aug 17 2021
#
#@author: Paul
#"""
import random
import numpy as np
import joblib
from itertools import combinations
import copy
from neorl.evolu.discrete import mutate_discrete, encode_grid_to_discrete
from neorl.evolu.discrete import decode_discrete_to_grid, encode_grid_indv_to_discrete
from neorl.utils.seeding import set_neorl_seed
from neorl.utils.tools import get_population, check_mixed_individual
[docs]class TS(object):
"""
Tabu Search Algorithm
:param mode: (str) problem type, either ``min`` for minimization problem or ``max`` for maximization
:param bounds: (dict) input parameter type and lower/upper bounds in dictionary form. Example: ``bounds={'x1': ['int', 1, 4], 'x2': ['int', -10, 10], 'x3': ['int', -100, 100]}``
:param fit: (function) the fitness function
:param tabu_tenure: (int): Timestep under which a certain list of solution cannot be accessed (for diversification). Default value is 6.
:param penalization_weight: (float): a scalar value for the coefficient that controls exploration/exploitation,
i.e. importance of the frequency of a certain action performed in the search. The higher the value, the least likely is an action to be performed again after
multiple attempts.
:param swap_mode: (str): either "swap" for swapping two elements of the input or "perturb" to perturb each input within certain bounds (see **Notes** below)
:param ncores: (int) number of parallel processors (only ``ncores=1`` is supported now)
:param seed: (int) random seed for sampling
"""
def __init__(self, mode, bounds, fit, tabu_tenure=6, penalization_weight = 0.8, swap_mode = "perturb", ncores=1, seed=None):
set_neorl_seed(seed)
assert ncores == 1,'-error: parallel implementaiton is not yet available. ncores ({}) should be equal to 1.'.format(ncores)
#assert ncores <= len(bounds), '--error: ncores ({}) must be less than or equal than the length of an individual solution ({})'.format(ncores, len(bounds))
int_transform='nearest_int'
self.mode=mode # mode for optimization: CS only solves a minimization problem.
if mode == 'min':
self.fit=fit
elif mode == 'max':
def fitness_wrapper(*args, **kwargs):
return -fit(*args, **kwargs)
self.fit=fitness_wrapper
else:
raise ValueError('--error: The mode entered by user is invalid, use either `min` or `max`')
self.int_transform=int_transform
if int_transform not in ["nearest_int", "sigmoid", "minmax"]:
raise ValueError('--error: int_transform entered by user is invalid, must be `nearest_int`, `sigmoid`, or `minmax`')
self.bounds=bounds
self.ncores = ncores
assert swap_mode in ["swap","perturb"],'--error: swap_mode must be either "swap" or "perturb" not ({})'.format(swap_mode)
self.swap_mode = swap_mode # swapping method for characterizing a "move" in the tabu search
self.tabu_tenure = tabu_tenure
self.penalization_weight = penalization_weight
self.ntabus = len(list(bounds.keys())) # the number of move performed depends on the size of the instance
self.var_type = np.array([bounds[item][0] for item in bounds])# infer variable types
#mir-grid
if "grid" in self.var_type:
self.grid_flag=True
self.orig_bounds=bounds #keep original bounds for decoding
#print('--debug: grid parameter type is found in the space')
self.bounds, self.bounds_map=encode_grid_to_discrete(self.bounds) #encoding grid to int
#define var_types again by converting grid to int
self.var_type = np.array([self.bounds[item][0] for item in self.bounds])
else:
self.grid_flag=False
self.bounds = bounds
self.orig_bounds=bounds
self.dim = len(bounds)
self.lb=np.array([self.bounds[item][1] for item in self.bounds])
self.ub=np.array([self.bounds[item][2] for item in self.bounds])
def init_sample(self, bounds):
#"""
#Initialize the initial population of Tabu
#"""
indv=[]
for key in bounds:
if bounds[key][0] == 'int':
indv.append(random.randint(bounds[key][1], bounds[key][2]))
elif bounds[key][0] == 'float':
indv.append(random.uniform(bounds[key][1], bounds[key][2]))
else:
raise Exception ('unknown data type is given, either int, float, or grid are allowed for parameter bounds')
return indv
def eval_tabus(self,tabu_list = None):
#---------------------
# Fitness calcs
#---------------------
core_lst=[]
if tabu_list is None:
for case in range (0, self.Positions.shape[0]):
core_lst.append(self.Positions[case, :])
if self.ncores > 1:
with joblib.Parallel(n_jobs=self.ncores) as parallel:
fitness_lst=parallel(joblib.delayed(self.fit_worker)(item) for item in core_lst)
else:
fitness_lst=[]
for item in core_lst:
fitness_lst.append(self.fit_worker(item))
else:
for case in range (0, tabu_list.shape[0]):
core_lst.append(tabu_list[case, :])
if self.ncores > 1:
with joblib.Parallel(n_jobs=self.ncores) as parallel:
fitness_lst=parallel(joblib.delayed(self.fit_worker)(item) for item in core_lst)
else:
fitness_lst=[]
for item in core_lst:
fitness_lst.append(self.fit_worker(item))
return fitness_lst
def select(self, pos, fit):
best_fit=np.min(fit)
min_idx=np.argmin(fit)
best_pos=pos[min_idx,:].copy()
return best_pos, best_fit
def ensure_bounds(self, vec):
vec_new = []
# cycle through each variable in vector
for i, (key, val) in enumerate(self.bounds.items()):
# variable exceedes the minimum boundary
if vec[i] < self.bounds[key][1]:
vec_new.append(self.bounds[key][1])
# variable exceedes the maximum boundary
if vec[i] > self.bounds[key][2]:
vec_new.append(self.bounds[key][2])
# the variable is fine
if self.bounds[key][1] <= vec[i] <= self.bounds[key][2]:
vec_new.append(vec[i])
return vec_new
def fit_worker(self, x):
#This worker is for parallel calculations
# Clip the tabu with position outside the lower/upper bounds and return same position
x=self.ensure_bounds(x)
if self.grid_flag:
#decode the individual back to the int/float/grid mixed space
x=decode_discrete_to_grid(x,self.orig_bounds,self.bounds_map)
# Calculate objective function for each search agent
fitness = self.fit(x)
return fitness
def ensure_discrete(self, vec):
#"""
#to mutate a vector if discrete variables exist
#handy function to be used within CS phases
#Params:
#vec - tabu position in vector/list form
#Return:
#vec - updated tabu position vector with discrete values
#"""
for dim in range(self.dim):
if self.var_type[dim] == 'int':
vec[dim] = mutate_discrete(x_ij=vec[dim],
x_min=min(vec),
x_max=max(vec),
lb=self.lb[dim],
ub=self.ub[dim],
alpha=self.a,
method=self.int_transform,
)
return vec
def UpdateTabu(self,Position, i, j):
#"""
#Perturb the 'Position' vector with parameter i and j
#if swap option: swap elements
#if perturb option: perturb elements
#"""
tempposition = copy.deepcopy(Position)
if self.swap_mode == "swap":
# job index in the Position:
i_index = tempposition.index(i)
j_index = tempposition.index(j)
tempposition[i_index], tempposition[j_index] = tempposition[j_index], tempposition[i_index]# Swap
elif self.swap_mode == "perturb":
if self.var_type[i] == 'int':
tempposition[i] = random.randint(j[0],j[1])
elif self.var_type[i] == 'float':
tempposition[i] = random.uniform(j[0],j[1])
return tempposition
[docs] def evolute(self,ngen,x0=None, verbose=False):
"""
This function evolutes the TS algorithm for number of generations
:param ngen: (int) number of generations to evolute
:param x0: (list) initial position of the tabu (vector size must be of same size as ``len(bounds)``)
:param verbose: (bool) print statistics to screen
:return: (tuple) (best individual, best fitness, and dictionary containing major search results)
"""
self.history = {'local_fitness':[], 'global_fitness':[]}
self.best_fitness=float("inf")
self.verbose=verbose
self.Positions = np.zeros(self.dim)#np.zeros((self.ntabus, self.dim))
if x0:
if self.ncores==1:
if not any(isinstance(el, list) for el in x0): #ensure a list of list is submitted for ncores=1
x0=[x0]
assert len(x0[0]) == self.ntabus, '--error: the length of individual in x0 ({}) MUST equal the size of the problem ({})'.format(len(x0[0]), self.ntabus)
assert len(x0) == self.ncores, '--error: the length of x0 ({}) MUST equal the number of ncores/parallel chains ({})'.format(len(x0), self.ncores)
for i in range(self.ncores):
check_mixed_individual(x=x0[i], bounds=self.orig_bounds) #assert the type provided is consistent
if self.grid_flag:
self.Positions = encode_grid_indv_to_discrete(x0[i], bounds=self.orig_bounds, bounds_map=self.bounds_map)
else:
self.Positions = x0[i]
else:
#self.Positions=self.init_sample(self.bounds) #TODO, update later for mixed-integer optimisation
# Initialize the positions of tabu
if self.swap_mode == "swap":# no repetition of element if swap method
self.Positions = list(range(1, self.ntabus + 1))
random.shuffle(self.Positions)
elif self.swap_mode == "perturb":
self.Positions = self.init_sample(self.bounds)
fitness=self.fit_worker(self.Positions) # evaluate the initial tabu
self.best_position, self.best_fitness = self.Positions.copy(), fitness#self.select(pos = self.Positions,fit = fitness) # find the initial best position and fitness
current_solution = self.Positions.copy()
tabu_structure = {} # record possible the tabu memory for all possible moves
if self.swap_mode == "swap":
for swap in combinations(range(1,self.ntabus + 1), 2):
tabu_structure[swap] = {'tabu_time': 0, 'MoveValue': 0, 'freq': 0, 'Penalized_MV': 0}
elif self.swap_mode == "perturb":
for l in range(self.ntabus):
tabu_structure[l] = {'tabu_time': 0, 'MoveValue': 0, 'freq': 0, 'Penalized_MV': 0}
iter = 1
for l in range(1, ngen+1):# Main loop
#-----------------------------
# Performs multiple moves and evaluate the resulting tabu
#-----------------------------
increment = 0 # utilize to perturb the Position vector at the 'increment'th position
temp_candidate = [] # store new candidate for perturbation to avoid the problem of change in perturbation when best move is called
for move in tabu_structure:# Searching the whole neighborhood of the current solution:
if self.swap_mode == "swap":
candidate_solution = self.UpdateTabu(self.Positions, move[0], move[1])
elif self.swap_mode == "perturb":
candidate_solution = self.UpdateTabu(self.Positions, increment,[self.lb[increment], self.ub[increment]])
temp_candidate.append(candidate_solution)
increment +=1
fitness = self.fit_worker(candidate_solution)
tabu_structure[move]['MoveValue'] = fitness
tabu_structure[move]['Penalized_MV'] = fitness + (tabu_structure[move]['freq'] *
self.penalization_weight)# Penalized fitness by simply adding freq to it (minimization):
#----------------------
# Manipulate the tabu list
#----------------------
while True:# Admissible move
best_move = min(tabu_structure, key =lambda x: tabu_structure[x]['Penalized_MV']) # select the move with the lowest Penalized fitness in the neighborhood (minimization)
MoveValue = tabu_structure[best_move]["MoveValue"]
tabu_time = tabu_structure[best_move]["tabu_time"]
if tabu_structure[best_move]['Penalized_MV'] > 1e11:# no improvement
break
if tabu_time < iter:# Not Tabu: the current move can be potentially added to the tabu list
# make the move
if self.swap_mode == "swap":
self.Positions = self.UpdateTabu(self.Positions, best_move[0], best_move[1])
elif self.swap_mode == "perturb":
best_loc = np.argmin([tabu_structure[x]['Penalized_MV'] for x in tabu_structure.keys()])
self.Positions = temp_candidate[best_loc].copy()
fitness = MoveValue#self.fit(self.Positions)
if MoveValue < self.best_fitness:# Best Improving move
self.best_position = self.Positions.copy()
self.best_fitness = fitness
# update tabu_time for the move and freq count
tabu_structure[best_move]['tabu_time'] = iter + self.tabu_tenure
tabu_structure[best_move]['freq'] += 1
iter += 1
break
else:# If tabu
# Aspiration
if MoveValue < self.best_fitness:
# make the move
if self.swap_mode == "swap":
self.Positions = self.UpdateTabu(self.Positions, best_move[0], best_move[1])
elif self.swap_mode == "perturb":
best_loc = np.argmin([tabu_structure[x]['Penalized_MV'] for x in tabu_structure.keys()])
self.Positions = temp_candidate[best_loc].copy()
fitness = self.fit_worker(self.Positions)
self.best_position = self.Positions.copy()
self.best_fitness = fitness
tabu_structure[best_move]['freq'] += 1
iter += 1
break
else:
tabu_structure[best_move]['Penalized_MV'] = float('inf')
continue
#----------------------
# Logger related portion
#----------------------
#for i, fits in enumerate(fitness):
for i, fits in enumerate([fitness]):
#save the best of the best!!!
if fits < self.best_fitness:
self.best_fitness=fits
#self.best_position=self.Positions[i, :].copy()
self.best_position=self.Positions.copy()
#--mir
if self.mode=='max':
self.fitness_best_correct=-self.best_fitness
self.local_fitness=-np.min(fitness)
else:
self.fitness_best_correct=self.best_fitness
self.local_fitness=np.min(fitness)
self.history['local_fitness'].append(self.local_fitness)
self.history['global_fitness'].append(self.fitness_best_correct)
# Print statistics
if self.verbose and l % self.ntabus:
print('^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^')
print('TS step {}/{}, ntabus={}, Ncores={}'.format((l)*self.ntabus, ngen*self.ntabus, self.ntabus, self.ncores))
print('^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^')
print('Best Tabu Fitness:', np.round(self.fitness_best_correct,6))
if self.grid_flag:
self.tabu_decoded = decode_discrete_to_grid(self.best_position, self.orig_bounds, self.bounds_map)
print('Best Tabu Position:', self.tabu_decoded)
else:
print('Best Tabu Position:', self.best_position)
print('^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^')
#mir-grid
if self.grid_flag:
self.tabu_correct = decode_discrete_to_grid(self.best_position, self.orig_bounds, self.bounds_map)
else:
self.tabu_correct = self.best_position
if self.verbose:
print('------------------------ TS Summary --------------------------')
print('Best fitness (y) found:', self.fitness_best_correct)
print('Best individual (x) found:', self.tabu_correct)
print('--------------------------------------------------------------')
return self.tabu_correct, self.fitness_best_correct, self.history