def __init__(self, instance_name: str, file_type: str = '.tsp'):
self.instance_name = instance_name
if file_type == '.tsp':
self.instance = tsplib95.load(Problem.files_path +
self.instance_name + file_type)
self.nx_graph = self.instance.get_graph()
try:
self.tour = tsplib95.load(Problem.files_path +
self.instance_name +
Problem.tour_files_suffix)
self.opt_cost = self.instance.trace_tours(self.tour.tours)[0]
self.opt_tour = self.tour.tours[0]
except:
self.tour = None
self.opt_cost = None
self.opt_tour = None
elif file_type == '.graphml':
self.instance = None
self.tour = None
self.nx_graph = nx.read_graphml(Problem.files_path +
self.instance_name + file_type)
self.opt_cost = 0
self.opt_tour = []
self.distance_matrix = nx.to_numpy_matrix(self.nx_graph)
self.instance = None
self.tour = None
self.size = self.distance_matrix.shape[0]
if self.instance_name == "brazil58":
self.opt_cost = 25395
def main():
problem_a: StandardProblem = tsplib95.load('data/kroa200.tsp')
problem_b: StandardProblem = tsplib95.load('data/krob200.tsp')
shutil.rmtree("./graphs/", ignore_errors=True)
os.makedirs('graphs/')
global_convexity_tests(problem_a, title="problem_a_node")
global_convexity_tests(problem_a,
similarity_function=edge_similarity,
title="problem_a_edge")
global_convexity_tests(problem_b, title="problem_b_node")
global_convexity_tests(problem_b,
similarity_function=edge_similarity,
title="problem_b_edge")
run_experiment_iterative_local_search(problem_a,
ImprovedHybridEvolutionarySolver(),
result_title="kroa200_ihe",
experiment_count=_COUNT,
max_time=20.777)
run_experiment_iterative_local_search(problem_b,
ImprovedHybridEvolutionarySolver(),
result_title="krob200_ihe",
experiment_count=_COUNT,
max_time=20.205)
def _create_training_data(self):
datadir = self.datadir
problems = []
graphs = []
tours = []
cur_nodes = []
adjacencies = []
for filename in os.listdir(datadir):
name, ext = os.path.splitext(filename)
tourpath = f'{name}.opt.tour'
if ext == '.tsp' and tourpath in os.listdir(datadir):
problem = tsp.load(os.path.join(datadir, filename))
if problem.type != 'TSP' or problem.edge_weight_type != 'EUC_2D' or problem.dimension < 20:
# Only using 2D symmetric graphs for TSP
# also no point in training if graph is so small (could always generate data elsewhere)
continue
print(problem.name, problem.dimension)
graph = []
for i in range(1, problem.dimension + 1):
graph.append(problem.node_coords[i])
adj = self.weighted_adj_from_pos(torch.tensor(graph))
tour = tsp.load(os.path.join(datadir, tourpath))
if np.array(tour.tours).shape[0] > 1:
print(tour.name)
breakpoint()
opttour = np.array(tour.tours[0])
for i in range(len(opttour)):
curtour = np.append(opttour[i:], opttour[:i])
curtour = curtour - 1 # adjust to 0 indexed numbering to match graph
curnode = curtour[0]
graphs.append(graph)
tours.append(curtour)
cur_nodes.append(curnode)
problems.append(problem.name)
adjacencies.append(adj)
opttour = opttour[::-1] # repeat for reversed tour
for i in range(len(opttour)):
curtour = np.append(opttour[i:], opttour[:i])
curtour = curtour - 1 # adjust to 0 indexed numbering to match graph
curnode = curtour[0]
graphs.append(graph)
tours.append(curtour)
cur_nodes.append(curnode)
problems.append(problem.name)
adjacencies.append(adj)
return np.array(problems), np.array(graphs), np.array(tours), np.array(
cur_nodes), adjacencies
def load_problem(self, problem_filename):
"""
Load problem from file, as well as opt.tour if available
"""
self.problem = tsp.load(problem_filename)
CandidateSolution.set_problem(self.problem)
self.optimum = None
opt_tour = problem_filename[:-4] + ".opt.tour"
try:
self.optimum = CandidateSolution(tsp.load(opt_tour).tours[0])
except FileNotFoundError as err:
print("FileNotFoundError: {0}".format(err))
else:
pass
def main():
problem_a: StandardProblem = tsplib95.load('data/kroa100.tsp')
problem_b: StandardProblem = tsplib95.load('data/krob100.tsp')
shutil.rmtree("./graphs/", ignore_errors=True)
average_time = []
average_time.append(
run_experiment_local_search(problem_a, NodeSwapSteepSearch(),
"kroa100_nsss", _EXPERIMENT_COUNT))
average_time.append(
run_experiment_local_search(problem_a, EdgeSwapSteepSearch(),
"kroa100_esss", _EXPERIMENT_COUNT))
average_time.append(
run_experiment_local_search(problem_a,
GreedyLocalSearch(use_node_swap=True),
"kroa100_gls_ns", _EXPERIMENT_COUNT))
average_time.append(
run_experiment_local_search(problem_a,
GreedyLocalSearch(use_edge_swap=True),
"kroa100_gls_es", _EXPERIMENT_COUNT))
average_time.append(
run_experiment_local_search(problem_b, NodeSwapSteepSearch(),
"krob100_nsss", _EXPERIMENT_COUNT))
average_time.append(
run_experiment_local_search(problem_b, EdgeSwapSteepSearch(),
"krob100_esss", _EXPERIMENT_COUNT))
average_time.append(
run_experiment_local_search(problem_b,
GreedyLocalSearch(use_node_swap=True),
"krob100_gls_ns", _EXPERIMENT_COUNT))
average_time.append(
run_experiment_local_search(problem_b,
GreedyLocalSearch(use_edge_swap=True),
"krob100_gls_es", _EXPERIMENT_COUNT))
max_avg_time = max(average_time)
print(f"Max average time : {round(max_avg_time * 1000.0)}")
print()
run_experiment_local_search(problem_a, RandomSearch(max_avg_time),
"kroa100_rs", _EXPERIMENT_COUNT)
run_experiment_local_search(problem_b, RandomSearch(max_avg_time),
"krob100_rs", _EXPERIMENT_COUNT)
def main():
problem_a: StandardProblem = tsplib95.load('data/kroa200.tsp')
problem_b: StandardProblem = tsplib95.load('data/krob200.tsp')
shutil.rmtree("./graphs/", ignore_errors=True)
run_experiment_iterative_local_search(problem_a,
HybridEvolutionarySolver(),
"kroa200_he", _EXPERIMENT_COUNT,
20.777)
run_experiment_iterative_local_search(problem_b,
HybridEvolutionarySolver(),
"krob200_he", _EXPERIMENT_COUNT,
20.205)
def christofides(path):
problem = tsplib95.load(path)
G = problem.get_graph()
preproc(G)
print("preproc done")
T = kruskal(G)
print("kruskal done")
O = oddsubgraph(T, G)
print("oddsubgraph done")
M = matching(O)
print("matching done")
H = nx.MultiGraph()
H.add_edges_from(T.edges)
H.add_edges_from(M.edges)
eulerkreis = get_euler_kreis(H)
print("eulerkreis found")
hamiltonkreis = get_hamilton_kreis(eulerkreis)
print("hamiltonkreis found")
return weight(hamiltonian, G)
def load_problem(self, filename):
problem = tsplib95.load(filename)
coord = problem.as_name_dict().get('node_coords')
self._labels = list(coord.keys())
self._coordinates = list(coord.values())
self._distance_matrix = distance_matrix(self._coordinates,
tsplib95.distances.euclidean)
def tsplib_distance_matrix(tsplib_file: str) -> np.ndarray:
"""Distance matrix from a TSPLIB file
Parameters
----------
tsplib_file
A string with the complete path of the TSPLIB file (or just its name if
it is the in current path)
Returns
-------
distance_matrix
A ND-array with the equivalent distance matrix of the input file
Notes
-----
This function can handle any file supported by the `tsplib95` lib.
"""
tsp_problem = tsplib95.load(tsplib_file)
distance_matrix_flattened = np.array(
[tsp_problem.get_weight(*edge) for edge in tsp_problem.get_edges()])
distance_matrix = np.reshape(
distance_matrix_flattened,
(tsp_problem.dimension, tsp_problem.dimension),
)
# Some problems with EXPLICIT matrix have a large number in the distance
# from a node to itself, which makes no sense for our problems. Thus,
# always ensure a diagonal filled with zeros
np.fill_diagonal(distance_matrix, 0)
return distance_matrix
def main():
cities = tsplib95.load(filePath)
n_nodes = len(list(cities.get_nodes()))
solutions = []
best_round_value = (0, 0)
for i in range(iterations):
nodes = list(range(1, n_nodes + 1))
random.shuffle(nodes) #initial configuration|: sort all nodes randomly
value = evaluation_function(nodes, cities, n_nodes)
step = 0
result = Solution(i)
result_file.write("\t============== Round: " + str(i + 1) +
" ==============\n")
start_t = time.time()
iterations_t = []
while (True):
step += 1
child_nodes, child_value = operate(nodes, cities, n_nodes)
result_file.write("\t==== " + str(step) + " iteration -> f(n): " +
str(value) + "\n")
if child_value <= value and step <= 300:
value = child_value
nodes = child_nodes
else:
result.set(nodes, value, step)
solutions.append(result) #print best results in file
result_file.write("\n\t==== Best Solution -> " + str(step) +
" iterations: f(n): " + str(value))
result_file.write("\n\t==== - Elapsed time: " +
str(sum(iterations_t)) +
" | time / iterations: " +
str(sum(iterations_t) / step) + "\n\n\n")
if best_round_value[0] == 0 or value < best_round_value[0]:
best_round_value = (value, i + 1)
with open(result_file.name[:-4] + "_nodes.txt",
'w',
encoding='utf-8') as f: #print final node
f.write("\t===== Best Solution found, in round: " +
str(i + 1) + " -> f(n): " +
str(best_round_value[0]) + " ######\n")
f.write("\t===== - Elapsed time: " +
str(sum(iterations_t)) +
" | time / iterations: " +
str(sum(iterations_t) / step) + "\n")
f.write("Order to visit cities: ")
for n in nodes:
f.write(str(n) + ", ")
if i == step - 1:
result_file.write("Best Solution: " +
str(best_round_value[1]) +
" round -> f(n): " +
str(best_round_value[0]))
break
iterations_t.append(time.time() - start_t)
start_t = time.time()
result_file.close()
def main():
problem_a: StandardProblem = tsplib95.load('data/kroa100.tsp')
problem_b: StandardProblem = tsplib95.load('data/krob100.tsp')
shutil.rmtree("graphs/", ignore_errors=True)
run_experiment_constructive(problem_a, NearestNeighbourProblemSolver(),
"kroa100_nn", _EXPERIMENT_COUNT)
run_experiment_constructive(problem_b, NearestNeighbourProblemSolver(),
"krob100_nn", _EXPERIMENT_COUNT)
run_experiment_constructive(problem_a, GreedyCycleProblemSolver(),
"kroa100_gc", _EXPERIMENT_COUNT)
run_experiment_constructive(problem_b, GreedyCycleProblemSolver(),
"krob100_gc", _EXPERIMENT_COUNT)
run_experiment_constructive(problem_a, RegretCycleProblemSolver(),
"kroa100_rc", _EXPERIMENT_COUNT)
run_experiment_constructive(problem_b, RegretCycleProblemSolver(),
"krob100_rc", _EXPERIMENT_COUNT)
def main():
problem_a: StandardProblem = tsplib95.load('data/kroa200.tsp')
problem_b: StandardProblem = tsplib95.load('data/krob200.tsp')
shutil.rmtree("./graphs/", ignore_errors=True)
run_experiment_local_search(problem_a, CandidateSteepSearch(),
"kroa200_css", _EXPERIMENT_COUNT)
run_experiment_local_search(problem_b, CandidateSteepSearch(),
"krob200_css", _EXPERIMENT_COUNT)
# This algorithm is implemented using wrong method
# run_experiment_local_search(problem_a, ScoreSteepSearch(), "kroa200_sss", _EXPERIMENT_COUNT)
# run_experiment_local_search(problem_b, ScoreSteepSearch(), "krob200_sss", _EXPERIMENT_COUNT)
run_experiment_local_search(problem_a, EdgeSwapSteepSearch(),
"kroa200_ess", _EXPERIMENT_COUNT)
run_experiment_local_search(problem_b, EdgeSwapSteepSearch(),
"krob200_ess", _EXPERIMENT_COUNT)
run_experiment_constructive(problem_a, GreedyCycleProblemSolver(),
"kroa200_gc", _EXPERIMENT_COUNT)
run_experiment_constructive(problem_b, GreedyCycleProblemSolver(),
"krob200_gc", _EXPERIMENT_COUNT)
def run(filename, is_plot=True):
datapath = get_example_path()
filepath = os.path.join(datapath, filename)
problem = tsplib95.load(filepath)
start = time()
solver = TSPSolver.from_tspfile(filepath)
solution = solver.solve()
end = time()
print("Time spent to solve {}s".format(end - start))
print("Optimal value: ", solution.optimal_value)
tour = solution.tour
tour = [t + 1 for t in tour]
plot(problem, tour)
def solve(solver='LKH', problem=None, **params):
assert shutil.which(solver) is not None, f'{solver} not found.'
valid_problem = problem is not None and isinstance(
problem, tsplib.models.StandardProblem)
assert ('problem_file' in params
) ^ valid_problem, 'Specify a TSPLIB95 problem object *or* a path.'
if problem is not None:
# hack for bug in tsplib
if len(problem.depots) > 0:
problem.depots = map(lambda x: f'{x}\n', problem.depots)
prob_file = tempfile.NamedTemporaryFile(mode='w')
problem.write(prob_file)
prob_file.write('\n')
prob_file.flush()
params['problem_file'] = prob_file.name
# need dimension of problem to parse solution
problem = tsplib.load(params['problem_file'])
if 'tour_file' not in params:
tour_file = tempfile.NamedTemporaryFile(mode='w')
params['tour_file'] = tour_file.name
with tempfile.NamedTemporaryFile(mode='w') as par_file:
par_file.write('SPECIAL\n')
for k, v in params.items():
par_file.write(f'{k.upper()} = {v}\n')
par_file.flush()
try:
subprocess.check_output([solver, par_file.name],
stderr=subprocess.STDOUT)
except subprocess.CalledProcessError as e:
raise Exception(e.output.decode())
solution = tsplib.load(params['tour_file'])
return solution.tours
def tsp_input():
global problem
global ARC
global A
global Weighted_ARC
global Weighted_ARC_Sorted
global input_nodes
#Uploading of input data from following directory
roots = Tk.Tk()
roots.withdraw()
roots.filename = filedialog.askopenfilename(
initialdir="/",
title="Select file",
filetypes=(("tsp files", "*.tsp"), ("all files", "*.*")))
problem = tsplib95.load(roots.filename)
nodes = list(problem.get_nodes())
input_nodes = len(nodes)
coords = []
ARC = []
for i in nodes:
coords.append(problem.node_coords[i])
#Initialization of adjacency matrix
A = np.ones((len(nodes), len(nodes)))
A = np.triu(A)
for i in nodes:
A[i - 1][i - 1] = 0
Weighted_ARC = []
Weighted_ARC_Sorted = []
for i in nodes:
for j in nodes:
if A[i - 1][j - 1] == 1:
#Weight of each arc is the distance between the two extremal nodes
x = ((coords[j - 1][0] - coords[i - 1][0])**2 +
(coords[j - 1][1] - coords[i - 1][1])**2)**(1 / 2)
Weighted_ARC.append([i, j, x])
ARC.append(tuple([i, j]))
#Ordering of the arcs given by the weight
Weighted_ARC_Sorted = Weighted_ARC.copy()
Weighted_ARC_Sorted.sort(key=operator.itemgetter(2))
return
def generate_cities(self, filename):
problem = tsplib95.load(filename)
G = problem.get_graph()
nodes = list(problem.get_nodes())
outer_array = []
for i, o in enumerate(nodes):
for j, v in enumerate(nodes):
if i == j:
outer_array.append([o - 1, v - 1, 0])
else:
outer_array.append([o - 1, v - 1, G.edges[o, v]['weight']])
self.edges = outer_array
self.data = len(nodes)
self.graph = problem
def _read_opt_sol(path: str):
"""
Ler o arquivo com a solução ótima da tsplib
:param path: path para o arquivo
:return: set com as tuplas (i,j), onde i < j, correspondente as arestas da solução
"""
tour = tsplib.load(path)
tour = np.array(tour.tours[0]) - 1
edges = set([(tour[0], tour[-1])])
for i in range(1, len(tour)):
edges.add((tour[i - 1],
tour[i]) if tour[i - 1] < tour[i] else (tour[i],
tour[i - 1]))
return edges
def main():
problem_a: StandardProblem = tsplib95.load('data/kroa200.tsp')
problem_b: StandardProblem = tsplib95.load('data/krob200.tsp')
shutil.rmtree("./graphs/", ignore_errors=True)
avg_time_kroa = run_experiment_local_search(problem_a, MultipleStartLocalSearch(), "kroa200_msls",
_EXPERIMENT_COUNT)
avg_time_krob = run_experiment_local_search(problem_b, MultipleStartLocalSearch(), "krob200_msls",
_EXPERIMENT_COUNT)
run_experiment_iterative_local_search(problem_a, IteratedLocalSearch1(), "kroa200_ils1", _EXPERIMENT_COUNT,
avg_time_kroa)
run_experiment_iterative_local_search(problem_b, IteratedLocalSearch1(), "krob200_ils1", _EXPERIMENT_COUNT,
avg_time_krob)
run_experiment_iterative_local_search(problem_a, IteratedLocalSearch2(), "kroa200_ils2", _EXPERIMENT_COUNT,
avg_time_kroa)
run_experiment_iterative_local_search(problem_b, IteratedLocalSearch2(), "krob200_ils2", _EXPERIMENT_COUNT,
avg_time_krob)
run_experiment_iterative_local_search(problem_a, IteratedLocalSearch2a(), "kroa200_ils2a", _EXPERIMENT_COUNT,
avg_time_kroa)
run_experiment_iterative_local_search(problem_b, IteratedLocalSearch2a(), "krob200_ils2b", _EXPERIMENT_COUNT,
avg_time_krob)
def load_problems(n):
source = pathlib.Path.home(
) / 'Desktop/TSP/TSPLIB' #hardcoded source, change later
p_path = source / 'problems'
s_path = source / 'solutions'
problems = sorted(list(p_path.glob('*.tsp')))[:n]
solutions = sorted(list(s_path.glob('*.opt.tour')))[:n]
p_s = list(zip(problems, solutions))
problems = {
i.stem: {
'n_nodes': int(re.split(r'\D', i.stem)[-1]),
'problem': tsplib95.load(i),
'opt_path': tsplib95.load(j).tours,
}
for (i, j) in p_s
}
for k, v in problems.items():
pnodes = min(list(v['problem'].get_nodes()))
snodes = min(v['opt_path'][0])
if pnodes == 0 and snodes == 1:
problems[k]['opt_sol'] = v['problem'].trace_tours(
[[i - 1 for i in v['opt_path'][0]]])[0]
elif (pnodes == 1 and snodes == 1) or (pnodes == 0 and snodes == 0):
problems[k]['opt_sol'] = v['problem'].trace_tours(
[v['opt_path'][0]])[0]
elif pnodes == 1 and snodes == 0:
problems[k]['opt_sol'] = v['problem'].trace_tours(
[[i + 1 for i in v['opt_path'][0]]])[0]
else:
raise ValueError("Problem nodes have abnormal indices")
return problems
def read_graphs(path_dir: str):
"""
Ler todos arquivo de instancias da tsplib no diretório
:param path_dir: path para o diretório contendo arquivos de isntâncias da tsplib
:return: dicionários {<nome da instância>: <nx graph>}
"""
g = {}
file_list = os.listdir(path_dir)
for i, file_name in enumerate(file_list):
g[os.path.basename(file_name).split('.')[0]] = tsplib.load(
os.path.join(path_dir, file_name)).get_graph(normalize=True)
progress(i + 1, len(file_list), 'Lendo instâncias', file_name)
# break
print()
return g
def getDictFromTspFile(tspFile):
p = tsp.load(tspFile)
if not p.is_depictable:
printf("Problem is not depictable!")
# Amendments as we need the nodes lexographically ordered
nnodes = len(list(p.get_nodes()))
i = 0
while nnodes > 1:
nnodes = nnodes / 10
i += 1
formatString = f"{{:0{i}d}}"
nodes = {
formatString.format(value): p.node_coords[index + 1]
for index, value in enumerate(p.get_nodes())
}
return nodes
def reset(self):
self.problem = tsplib95.load(self.tsp_name)
self.node_position = self.problem.as_name_dict().get("node_coords")
#Construct a Graph
self.node_from = []
self.node_to = []
self.node_features = []
for nodeID, position in self.node_position.items():
adjacent_nodes = []
node_feature = position + [0 for _ in range(12)]
for _ in range(12):
adjacent_nodeID = random.sample(
list(self.node_position.keys()), 1)[0]
while adjacent_nodeID == nodeID:
adjacent_nodeID = random.sample(
list(self.node_position.keys()), 1)[0]
adjacent_nodes.append(adjacent_nodeID)
adjacent_nodes = sorted(adjacent_nodes)
for i in range(12):
self.node_from.append(nodeID - 1)
self.node_to.append(adjacent_nodes[i] - 1)
for i in range(12):
node_feature[i + 2] = self.statistics[nodeID].get(i)
self.node_features.append(node_feature)
self.graph = dgl.DGLGraph((self.node_from, self.node_to))
self.graph.ndata["h"] = np.float32(self.node_features)
self.graph.edata["h"] = np.float32(
np.random.randn(len(self.node_position) * 12, 5))
#Randomly choose a starting point
self.start = random.randint(0, self.numberOfNodes - 1)
self.start_position = self.node_position.get(self.start + 1)
return self.graph, self.start
def index(file_name):
cities = tsplib95.load(
os.path.join(os.getcwd(), '../data/' + file_name + '.tsp'))
initial_state = random_state(cities)
distance_initial_state = calc_distance_of_state(initial_state, cities)
best_distance, best_solution = swap_neighbours(initial_state, cities)
print(initial_state)
plot_path(cities, initial_state, distance_initial_state, 'estado_inicial',
file_name)
print('initial: {}'.format(distance_initial_state))
print(best_solution)
plot_path(cities, best_solution, best_distance, 'melhor_solucao',
file_name)
print('best: {}'.format(best_distance))
print('difference: {}'.format(distance_initial_state - best_distance))
def load_file_and_create_graph(self, filename):
problem = tsplib95.load(filename)
G = problem.get_graph()
nodes = list(problem.get_nodes())
outer_array = []
for i, o in enumerate(nodes):
inner_array = []
for j, v in enumerate(nodes):
if i == j:
inner_array.append([0, 1])
elif i > j:
inner_array.append(outer_array[j][i])
else:
inner_array.append([G.edges[o, v]['weight'], 1])
outer_array.append(inner_array)
self.edges = outer_array
self.cities = self.ants = len(nodes)
self.nodes = nodes
self.graph = problem
def main_run():
global cm, coords, WIDTH, HEIGHT, CITIES
path = os.path.join(os.path.dirname(__file__), 'data/fri26.tsp')
problem = tsplib95.load(path)
print(list(problem.get_nodes()))
CITIES = len(list(problem.get_nodes()))
for i in range(0, CITIES):
for j in range(0, CITIES):
edge = i, j
weight = problem.get_weight(*edge)
cm[i, j] = weight
genome = G1DList.G1DList(CITIES)
genome.setParams(dist=cm)
genome.evaluator.set(lambda chromosome: tour_length(cm, chromosome))
genome.crossover.set(G1DListCrossoverPMX)
genome.mutator.set(G1DListMutatorDisplacement)
genome.initializator.set(G1DListTSPInitializatorRandom)
# 3662.69
ga = GSimpleGA.GSimpleGA(genome)
ga.setGenerations(GENERATION_COUNT)
ga.setMinimax(Consts.minimaxType["minimize"])
ga.setCrossoverRate(1.0)
ga.setMutationRate(0.02)
ga.setPopulationSize(80)
ga.selector.set(SelectionRank.SelectorLinearRanking)
# This is to make a video
if PIL_SUPPORT:
ga.stepCallback.set(evolve_callback)
# 21666.49
ga.evolve(freq_stats=1)
best = ga.bestIndividual()
if PIL_SUPPORT:
write_tour_to_img(coords, best, f"{RESULTS_DIRECTORY}/tsp_result.png")
else:
print("No PIL detected, cannot plot the graph !")
def christofides(path):
problem = tsplib95.load(path)
G = problem.get_graph()
preproc(G)
T = minspantree(G)
O = oddsubgraph(T, G)
M = perfectmatching(O)
H = nx.MultiGraph()
H.add_edges_from(T.edges)
H.add_edges_from(M.edges)
print(H.edges)
nx.draw_networkx(H)
#plt.show()
eulerian = EulerianCircle(H)
hamiltonian = HamiltonianCircle(eulerian)
#draw(G, hamiltonian)
return weight(hamiltonian, G)
def __init__(self, tsp_name) :
self.problem = tsplib95.load(tsp_name)
#Create a Graph
node_position = self.problem.as_name_dict().get("node_coords")
edges, distances, node_position = self.construct_edges(node_position)
self.edges = edges
self.distances = distances
self.node_position = node_position
node_from = []
node_to = []
for key, value in self.node_position.items() :
for i in range(8) :
node_from.append(key-1)
node_to.append(value[i+2]-1)
self.graph = dgl.DGLGraph((node_from, node_to))
self.graph.ndata["h"] = np.float32(list(node_position.values()))
self.graph.edata["h"] = np.float32(np.random.randn(len(self.node_position) * 8, 5))
self.visited = dict()
def __init__(self, tsp_name):
#Load a Problem
self.tsp_name = tsp_name
self.problem = tsplib95.load(self.tsp_name)
#Get Node Information
self.node_position = self.problem.as_name_dict().get("node_coords")
self.numberOfNodes = len(self.node_position)
#Get Distance Information
self.distances = dict()
self.statistics = dict()
for i in range(self.numberOfNodes):
self.distances[i + 1] = dict()
self.statistics[i + 1] = dict()
for j in range(12):
self.statistics[i + 1][j] = 0
for i in range(self.numberOfNodes):
for j in range(self.numberOfNodes):
if i >= j: continue
else:
distance = math.sqrt((self.node_position[i + 1][0] -
self.node_position[j + 1][0])**2 +
(self.node_position[i + 1][1] -
self.node_position[j + 1][1])**2)
self.distances[i + 1][j + 1] = distance
self.distances[j + 1][i + 1] = distance
distance_class = int(distance * (10**-3))
self.statistics[i + 1][min(
distance_class, 11)] = self.statistics[i + 1].get(
min(distance_class, 11)) + 1
self.statistics[j + 1][min(
distance_class, 11)] = self.statistics[j + 1].get(
min(distance_class, 11)) + 1
def main():
points = tsplib95.load('data/' + points_path)
n_nodes = len(list(points.get_nodes()))
all_value = 1000000000
file_name = points_path.split('.')[0] + '_' + swap_strategy + '_' + str(
n_tries)
if swap_strategy == 'random':
file_name += '_' + str(n_random_swap) + '_' + str(n_child_random_swap)
file_name += '_random_decay.txt'
file_log = open('results/log_' + file_name, 'w')
for i in tqdm.tqdm(range(n_tries)):
best_path, best_value, best_n = find_best_solution(points, n_nodes)
if best_value < all_value:
all_path, all_value, all_n, all_try = best_path, best_value, best_n, (
i + 1)
file_log.write("[" + str(i + 1) + "]:" + "\n")
file_log.write("\tDistância do melhor caminho: " + str(best_value) +
"\n")
file_log.write("\tIterações executadas: " + str(best_n) + "\n")
file_log.write("\tMelhor caminho: ")
write_list_file(file_log, best_path)
file_log.close()
file_best = open('results/best_' + file_name, 'w')
file_best.write("Overall best[" + str(all_try) + "]:" + "\n")
file_best.write("\tDistância do melhor caminho: " + str(all_value) + "\n")
file_best.write("\tIterações executadas: " + str(all_n) + "\n")
file_best.write("\tMelhor caminho: ")
write_list_file(file_best, all_path)
file_best.close()
import tsplib95
import acopy
# setting up the environment
solver = acopy.Solver(rho=.03, q=1)
colony = acopy.Colony(alpha=1, beta=3)
# adding a timer to record the total time to find the best path
timer = acopy.plugins.Timer()
solver.add_plugin(timer)
# setting it up so that is prints out each iteration
printout = acopy.plugins.Printout()
solver.add_plugin(printout)
problem = tsplib95.load('problems/pr124.tsp.txt') # loading in the dataset
G = problem.get_graph() # changing the dataset so that it can be used by the algorithm
tour = solver.solve(G, colony, limit=100) #running the algorithm
# printing out relevant information
print("Shortest tour: ", tour.cost)
print("Best tour: ", tour.nodes)
print("Time to complete: ", timer.duration)
