def main():
bench = Benchmark()
for num_function in range(INITIAL_FUNCTION, LAST_FUNCTION + 1):
info = bench.get_info(num_function)
print(f'\nFunction {num_function}: {info}')
for i in range(INITIAL_EJECUTION, LAST_EJECUTION + 1):
BKS = info['best']
D = info['dimension']
NP = 30
N_Gen = 5000
A = 0.95
r = 0.1
alpha = 0.9
gamma = 0.5
fmin = 0
fmax = 1
Lower = info['lower']
Upper = info['upper']
ObjetiveFunction = bench.get_function(num_function)
name_ejecution_file = f'function{num_function}_{i}.csv'
name_logs_file = 'Logs/' + name_ejecution_file
name_cluster_file = 'Logs/clusters/' + name_ejecution_file
bats = BatAlgorithm(num_function, i, BKS, D, NP, N_Gen, A, r,
alpha, gamma, fmin, fmax, Lower, Upper,
ObjetiveFunction)
bats.move_bats(name_logs_file, name_cluster_file, ORIGINAL_MH, 100)
def CC_exe(Dim, func_num, NIND, Max_iteration, scale_range, groups, method):
bench = Benchmark()
function = bench.get_function(func_num)
name = 'f' + str(func_num)
print(name, 'Optimization with', method)
"""The next is DE optimization"""
best_indexes, best_obj_trace = DE.CC(Dim, NIND, Max_iteration, function, scale_range, groups)
help.write_obj_trace(name, method, best_obj_trace)
def DECC_CL_exe(Dim, func_num, NIND, Max_iteration, scale_range, groups_One, groups_Lasso, Lasso_cost, candidate, method):
bench = Benchmark()
function = bench.get_function(func_num)
EFs = 3000000
name = 'f' + str(func_num)
print(name, 'Optimization with', method)
"""The next is DE optimization"""
best_indexes, best_obj_trace_CC, Population, up, down, cost = DE.DECC_CL_CCDE(Dim, NIND, Max_iteration, function,
scale_range, groups_One, candidate)
def go_lsgo(func_id):
from cec2013lsgo.cec2013 import Benchmark
bench = Benchmark()
bench_info = bench.get_info(func_id)
dim = bench_info['dimension']
return bench.get_function(func_id), None, [bench_info['lower']] * dim, [
bench_info['upper']
] * dim
def DECC_CL_exe(Dim, func_num, NIND, scale_range, groups_One, groups_Lasso, Lasso_cost, method):
bench = Benchmark()
function = bench.get_function(func_num)
EFs = 3000000
name = 'f' + str(func_num)
print(name, 'Optimization with', method)
"""The next is DE optimization"""
best_indexes, best_obj_trace_CC, cost, up, down = DE.DECC_CL_CCDE(Dim, NIND, int((EFs - Lasso_cost) / (NIND * Dim)),
function, scale_range, groups_One)
central_point = best_indexes[len(best_indexes)-1]
best_indexes, best_obj_trace_CL = DE.DECC_CL_DECC_L(Dim, NIND, int((EFs-Lasso_cost-cost)/(NIND*Dim)), function, up, down, groups_Lasso, central_point)
help.write_obj_trace(name, method, best_obj_trace_CC+best_obj_trace_CL)
def DECC_D(func_num, groups_num=10, max_number=100):
bench = Benchmark()
function = bench.get_function(func_num)
benchmark_summary = bench.get_info(func_num)
scale_range = [benchmark_summary['lower'], benchmark_summary['upper']]
Dim = 1000
groups = k_s(10, max_number)
delta = [0] * Dim
NIND = 10000
for i in range(len(groups)):
delta[i * max_number:(i + 1) * max_number] = DE.OptTool(
Dim, NIND, 1, function, groups[i], scale_range, -1)
sort_index = np.argsort(delta).tolist()
groups = []
for i in range(groups_num):
groups.append(sort_index[i * max_number:(i + 1) * max_number])
return groups
def DECC_DG(func_num):
cost = 2
bench = Benchmark()
function = bench.get_function(func_num)
groups = CCDE(1000)
intercept = function(np.zeros((1, 1000))[0])
for i in range(len(groups) - 1):
if i < len(groups) - 1:
cost += 2
index1 = np.zeros((1, 1000))[0]
index1[groups[i][0]] = 1
delta1 = function(index1) - intercept
for j in range(i + 1, len(groups)):
cost += 2
if i < len(groups) - 1 and j < len(
groups) and not help.DG_Differential(
groups[i][0], groups[j][0], delta1, function,
intercept):
groups[i].extend(groups.pop(j))
j -= 1
return groups, cost
checkpoints = (
np.array([0.04, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]) *
maxFEs).astype(int) #array
for fid in Functions:
#print("Processing f" + str(fid))
filename = "myresults_" + str(popsize1) + "pop1_" + str(
popsize2) + "pop2_f" + str(fid) + ".csv"
outputfile = open(filename, "w+")
outputfile.write("FEs,Value,Seed,Function\n")
for Rseed in seeds:
'''
Set objective function: the CEC'13 LSGO Benchmark presents 15 minimization problems.
'''
bench = Benchmark()
fobj = bench.get_function(fid) #objective function
info = bench.get_info(
fid
) #info is a dictionary which keys are lower, upper, dimension, threshold and best.
'''
Set parameters
'''
np.random.seed(Rseed)
currentFEs = 0
results = [
] #list of tuples. Each tuple will be of the form (checkpoint,decision_vector,value_of_decision_vector)
'''
Set components
'''
DE1 = shade(fobj,
info,
def main(args):
global SR_MTS, SR_global_MTS
"""
Main program. It uses
Run DE for experiments. F, CR must be float, or 'n' as a normal
"""
description = "SHADEILS"
parser = argparse.ArgumentParser(description)
parser.add_argument("-f",
default=16,
type=int,
dest="function",
help='function')
# Argument for problem type: D4, D12 or D19, both with or without noise
parser.add_argument("-pr",
required=True,
type=str,
choices=list(
("D4", "D12", "D19", "D4N", "D12N", "D19N")),
dest="problem",
help='problem type')
parser.add_argument("-v",
default=False,
dest="verbose",
action='store_true',
help='verbose mode')
parser.add_argument("-s",
default=1,
type=int,
dest="seed",
choices=range(1, 6),
help='seed (1 - 5)')
parser.add_argument("-r", default=1, type=int, dest="run", help='runs')
parser.add_argument("-e",
required=False,
type=int,
dest="maxevals",
help='maxevals')
# Argument for index of evals, that is [1,2,3] -> [1.2e5, 6e5,3e6]
parser.add_argument("-i",
default=1,
type=int,
dest="index",
choices=range(1, 4),
help="index of evals")
parser.add_argument("-t",
default=0.001,
type=float,
dest="threshold",
help='threshold')
parser.add_argument("-p",
default=100,
type=int,
dest="popsize",
help='population size')
parser.add_argument("-H",
default=3,
type=int,
dest="shade_h",
help='SHADE history size')
parser.add_argument("-d",
default="results",
type=str,
dest="dir_output",
help='directory output')
# Time stamp for comparative study
time0 = time.clock()
#seeds
seeds = [24, 45689, 97232447, 96793335, 12345679]
args = parser.parse_args(args)
fun = args.function
problem = args.problem
# Noise problem, False by default
noise = False
if (problem in ["D4", "D4N"]):
dim = 1024
elif (problem in ["D12", "D12N"]):
dim = 3072
elif (problem in ["D19", "D19N"]):
dim = 4864
if ("N" in problem):
print("Problem with noise")
noise = True
print("EEG Problem: {0}".format(problem))
print("Problem: {0}".format(problem))
print("Dimension: {0}".format(dim))
print("Seed: {0}".format(args.seed))
print("Treshold: {0}".format(args.threshold))
print("Popsize: {0}".format(args.popsize))
if args.shade_h is None:
args.shade_h = min(args.popsize, 100)
print("SHADE_H: {0}".format(args.shade_h))
if (args.maxevals):
evals = list(map(int, [1.0e5, 6e5, 3e6])[:args.maxevals])
else:
evals = list(map(int, [1.0e5, 6e5, 3e6]))
evals_index = args.index
bench = Benchmark()
maxfuns = bench.get_num_functions()
funinfo = bench.get_info(fun, dim)
if not (1 <= fun <= maxfuns and 1 <= args.seed <= 5):
parser.print_help()
sys.exit(1)
name = "SHADEILS"
fname = name + "_EEGProblem_" + problem + "_" + str(
evals[evals_index - 1]) + ".txt"
output = path.join(args.dir_output, fname)
if not args.verbose and isfile(output):
fin = open(output, 'rb')
lines = fin.readlines()
fin.close()
if lines:
print("Experiment already exists. Check results folder")
return
if not args.verbose:
fid = open(output, 'w')
else:
fid = sys.stdout
# Parameter commons
#bench.set_algname("shadeils_restart0.1_pos")
fitness_fun = bench.get_function(fun, dim, noise)
seed(seeds[args.seed - 1])
for _ in range(args.run):
SR_MTS = []
SR_global_MTS = []
SHADEILS.ihshadels(fitness_fun,
funinfo,
dim,
evals,
evals_index,
fid,
time0,
threshold=args.threshold,
popsize=args.popsize,
info_de=args.shade_h)
bench.next_run()
fid.close()
def main(args):
global SR_MTS, SR_global_MTS
"""
Main program. It uses
Run DE for experiments. F, CR must be float, or 'n' as a normal
"""
description = "IHDELS"
parser = argparse.ArgumentParser(description)
parser.add_argument("-f", required=True, type=int, choices=range(1, 16), dest="function", help='function')
parser.add_argument("-v", default=False, dest="verbose", action='store_true', help='verbose mode')
parser.add_argument("-s", default=1, type=int, dest="seed", choices=range(1, 6), help='seed (1 - 5)')
parser.add_argument("-r", default=5, type=int, dest="run", help='runs')
parser.add_argument("-e", required=False, type=int, dest="maxevals", help='maxevals')
parser.add_argument("-t", default=0.01, type=float, dest="threshold", help='threshold')
parser.add_argument("-p", default=100, type=int, dest="popsize", help='population size')
parser.add_argument("-H", default=None, type=int, dest="shade_h", help='SHADE history size')
parser.add_argument("-d", default="results", type=str, dest="dir_output", help='directory output')
#seeds
seeds = [23, 45689, 97232447, 96793335, 12345679]
args = parser.parse_args(args)
fun = args.function
dim = 1000
print("Function: {0}".format(fun))
print("Seed: {0}".format(args.seed))
print("Treshold: {0}".format(args.threshold))
print("Popsize: {0}".format(args.popsize))
if args.shade_h is None:
args.shade_h = min(args.popsize, 100)
print("SHADE_H: {0}".format(args.shade_h))
if (args.maxevals):
evals = list(map(int, [1.2e5, 6e5, 3e6])[:args.maxevals])
else:
evals = list(map(int, [1.2e5, 6e5, 3e6]))
print(evals)
bench = Benchmark()
maxfuns = bench.get_num_functions()
funinfo = bench.get_info(fun)
if not (1 <= fun <= maxfuns and 1 <= args.seed <= 5):
parser.print_help()
sys.exit(1)
name = "SHADEILS"
fname = name +"_pop{args.popsize}_H{args.shade_h}_t{args.threshold:.2f}_F{args.function}_{args.seed}r{args.run}.txt".format(args=args);
output = path.join(args.dir_output, fname)
if not args.verbose and isfile(output):
fin = open(output, 'rb')
lines = fin.readlines()
fin.close()
if lines:
print("Experiment already exists. Check results folder")
return
if not args.verbose:
fid = open(output, 'w')
else:
fid = sys.stdout
# Parameter commons
#bench.set_algname("shadeils_restart0.1_pos")
fitness_fun = bench.get_function(fun)
seed(seeds[args.seed-1])
for _ in range(args.run):
SR_MTS = []
SR_global_MTS = []
SHADEILS.ihshadels(fitness_fun, funinfo, dim, evals, fid, threshold=args.threshold, popsize=args.popsize, info_de=args.shade_h)
bench.next_run()
fid.close()
def LASSOCC(func_num):
Dim = 1000
group_dim = 20
size = group_dim * 100
degree = 3
bench = Benchmark()
max_variables_num = group_dim
function = bench.get_function(func_num)
benchmark_summary = bench.get_info(func_num)
scale_range = [benchmark_summary['lower'], benchmark_summary['upper']]
verify_time = 0
All_groups = []
intercept = function(np.zeros((1, 1000))[0])
one_bias = []
for i in range(1000):
index = np.zeros((1, 1000))[0]
index[i] = 1
one_bias.append(function(index) - intercept)
verify_time += 1001
for current_index in range(0, int(Dim / group_dim)):
# print(current_index)
Lasso_model, Feature_names = SparseModel.Regression(
degree, size, Dim, group_dim, current_index, scale_range, function)
# Grouping
coef, Feature_names = help.not_zero_feature(
Lasso_model.coef_, help.feature_names_normalization(Feature_names))
groups = help.group_DFS(group_dim, Feature_names, max_variables_num)
# print(groups)
bias = current_index * group_dim
for g in groups:
for i in range(len(g)):
g[i] += bias
for g in groups:
if not g or g is None:
groups.remove(g)
# We need to check the relationship between new groups and previous groups
temp_groups = []
for i in range(len(All_groups)):
for j in range(len(groups)):
if i < len(All_groups) and j < len(groups):
verify_time += 1
if not help.Differential(All_groups[i][0], groups[j][0],
function, intercept, one_bias):
g1 = All_groups.pop(i)
g2 = groups.pop(j)
temp_groups.append(g1 + g2)
i -= 1
j -= 1
break
for g in All_groups:
temp_groups.append(g)
for g in groups:
temp_groups.append(g)
All_groups = temp_groups.copy()
return All_groups, verify_time + 100000
from cec2013lsgo.cec2013 import Benchmark
import os.path as path
if __name__ == '__main__':
# func_num = 11
Dim = 1000
NIND = 30
bench = Benchmark()
EFs = 3000000
for func_num in range(2, 16):
test_time = 1
name = 'f' + str(func_num)
benchmark_summary = bench.get_info(func_num)
scale_range = [benchmark_summary['lower'], benchmark_summary['upper']]
function = bench.get_function(func_num)
groups_One = CCDE(Dim)
for iteration in range(1, 2):
EFs = 3000000
name = 'f' + str(func_num)
"""The next is DE optimization"""
best_indexes, best_obj_trace_CC, up, down = DE.DECC_CL_CCDE(Dim, NIND, iteration, function, scale_range, groups_One)
delta = []
for i in range(len(up)):
delta.append((up[i]-down[i])/(scale_range[1]-scale_range[0]))
up_bound = max(delta)
down_bound = min(delta)
mean = sum(delta) / len(delta)
