def test_benchmark_dag(self):
"""
Tests that random graph generators are working properly, and generates the
benchmark data that is saved to `benchmark.txt` in the root directory for
this project.
"""
# |V| = 4
pre_time_dag_one = time.time()
output_dag_one = DAGSolver(self.matrix_input_dag_one).topological_sort()
post_time_dag_one = time.time()
solving_time_dag_one = post_time_dag_one - pre_time_dag_one
self.assertIsNotNone(output_dag_one)
score_dag_one = scoreSolution(self.matrix_input_dag_one, output_dag_one)
file_string_one = "DAG, |V| = 4: Score = %.4f, Time = %.4f\n" % (score_dag_one, solving_time_dag_one)
self.benchmark_file.write(file_string_one)
print "Finished Benchmark 1, DAG"
# |V| = 7
pre_time_dag_two = time.time()
output_dag_two = DAGSolver(self.matrix_input_dag_two).topological_sort()
post_time_dag_two = time.time()
solving_time_dag_two = post_time_dag_two - pre_time_dag_two
self.assertIsNotNone(output_dag_two)
score_dag_two = scoreSolution(self.matrix_input_dag_two, output_dag_two)
file_string_two = "DAG, |V| = 7: Score = %.4f, Time = %.4f\n" % (score_dag_two, solving_time_dag_two)
self.benchmark_file.write(file_string_two)
print "Finished Benchmark 2, DAG"
# |V| = 20
pre_time_dag_three = time.time()
output_dag_three = DAGSolver(self.matrix_input_dag_three).topological_sort()
post_time_dag_three = time.time()
solving_time_dag_three = post_time_dag_three - pre_time_dag_three
self.assertIsNotNone(output_dag_three)
score_dag_three = scoreSolution(self.matrix_input_dag_three, output_dag_three)
file_string_three = "DAG, |V| = 20: Score = %.4f, Time = %.4f\n" % (score_dag_three, solving_time_dag_three)
self.benchmark_file.write(file_string_three)
print "Finished Benchmark 3, DAG"
# |V| = 100
pre_time_dag_four = time.time()
output_dag_four = DAGSolver(self.matrix_input_dag_four).topological_sort()
post_time_dag_four = time.time()
solving_time_dag_four = post_time_dag_four - pre_time_dag_four
self.assertIsNotNone(output_dag_four)
score_dag_four = scoreSolution(self.matrix_input_dag_four, output_dag_four)
file_string_four = "DAG, |V| = 100: Score = %.4f, Time = %.4f\n" % (score_dag_four, solving_time_dag_four)
self.benchmark_file.write(file_string_four)
print "Finished Benchmark 4, DAG"
# |V| = 1000
pre_time_dag_five = time.time()
output_dag_five = DAGSolver(self.matrix_input_dag_five).topological_sort()
post_time_dag_five = time.time()
solving_time_dag_five = post_time_dag_five - pre_time_dag_five
self.assertIsNotNone(output_dag_five)
score_dag_five = scoreSolution(self.matrix_input_dag_five, output_dag_five)
file_string_five = "DAG, |V| = 1000: Score = %.4f, Time = %.4f\n" % (score_dag_five, solving_time_dag_five)
self.benchmark_file.write(file_string_five)
print "Finished Benchmark 5, DAG"
def test_simulated_annealing(self):
random.seed(170)
observed_output_new_graph = SimulatedAnnealingSolver(
self.initial_ordering_new_graph,
self.matrix_new_graph
).maximum_acyclic_subgraph()
expected_score_new_graph = 54.0
observed_score_new_graph = scoreSolution(self.matrix_new_graph, observed_output_new_graph)
self.assertEquals(observed_score_new_graph, expected_score_new_graph)
observed_output_foster = SimulatedAnnealingSolver(
self.initial_ordering_foster,
self.matrix_foster
).maximum_acyclic_subgraph()
expected_score_foster = 113.0
observed_score_foster = scoreSolution(self.matrix_foster, observed_output_foster)
self.assertEquals(observed_score_foster, expected_score_foster)
expected_output_input_one = [0, 2, 1, 3]
observed_output_input_one = SimulatedAnnealingSolver(
self.initial_ordering_input_one,
self.matrix_input_one
).maximum_acyclic_subgraph()
self.assertEquals(observed_output_input_one, expected_output_input_one)
expected_ouput_input_cycle = [0, 1, 2, 3]
observed_output_input_cycle = SimulatedAnnealingSolver(
self.initial_ordering_input_cycle,
self.matrix_input_cycle
).maximum_acyclic_subgraph()
self.assertEquals(observed_output_input_cycle, expected_ouput_input_cycle)
expected_output_nonplanar_one = [7, 8, 9, 2, 12, 13, 3, 10, 11, 4, 5, 0, 1, 6]
observed_output_nonplanar_one = SimulatedAnnealingSolver(
self.initial_ordering_input_nonplanar_one,
self.matrix_input_nonplanar_one
).maximum_acyclic_subgraph()
self.assertEquals(observed_output_nonplanar_one, expected_output_nonplanar_one)
expected_output_nonplanar_two = [1, 5, 7, 2, 0, 6, 4, 3]
observed_output_nonplanar_two = SimulatedAnnealingSolver(
self.initial_ordering_input_nonplanar_two,
self.matrix_input_nonplanar_two
).maximum_acyclic_subgraph()
self.assertEquals(observed_output_nonplanar_two, expected_output_nonplanar_two)
observed_output_gray = SimulatedAnnealingSolver(
self.initial_ordering_input_gray,
self.matrix_input_gray
).maximum_acyclic_subgraph()
expected_score_gray = 70.0
observed_score_gray = scoreSolution(self.matrix_input_gray, observed_output_gray)
self.assertEquals(observed_score_gray, expected_score_gray)
def test_benchmark_feedback_arc_set_eades(self):
# |V| = 4
pre_time_eades_one = time.time()
output_solver_one = FinalSolver(self.matrix_input_brute_force_two)
output_eades_one = output_solver_one.obtain_library_solution(output_solver_one.adj_matrix, force_eades=True)
post_time_eades_one = time.time()
solving_time_eades_one = post_time_eades_one - pre_time_eades_one
self.assertIsNotNone(output_eades_one)
score_eades_one = scoreSolution(self.matrix_input_brute_force_two, output_eades_one)
file_string_one = "Eades Approximation, |V| = 4: Score = %.4f, Time = %.4f\n" % (
score_eades_one,
solving_time_eades_one,
)
self.benchmark_file.write(file_string_one)
# |V| = 25
pre_time_eades_two = time.time()
output_solver_two = FinalSolver(self.matrix_input_brute_force_five)
output_eades_two = output_solver_two.obtain_library_solution(output_solver_two.adj_matrix)
post_time_eades_two = time.time()
solving_time_eades_two = post_time_eades_two - pre_time_eades_two
self.assertIsNotNone(output_eades_two)
score_eades_two = scoreSolution(self.matrix_input_brute_force_five, output_eades_two)
file_string_two = "Eades Approximation, |V| = 25: Score = %.4f, Time = %.4f\n" % (
score_eades_two,
solving_time_eades_two,
)
self.benchmark_file.write(file_string_two)
# |V| = 100
pre_time_eades_three = time.time()
output_solver_three = FinalSolver(self.matrix_input_size_100)
output_eades_three = output_solver_three.obtain_library_solution(output_solver_three.adj_matrix)
post_time_eades_three = time.time()
solving_time_eades_three = post_time_eades_three - pre_time_eades_three
self.assertIsNotNone(output_eades_three)
score_eades_three = scoreSolution(self.matrix_input_size_100, output_eades_three)
file_string_three = "Eades Approximation, |V| = 100: Score = %.4f, Time = %.4f\n" % (
score_eades_three,
solving_time_eades_three,
)
self.benchmark_file.write(file_string_three)
# |V| = 1000
pre_time_eades_four = time.time()
output_solver_four = FinalSolver(self.matrix_input_size_1000)
output_eades_four = output_solver_four.obtain_library_solution(output_solver_four.adj_matrix)
post_time_eades_four = time.time()
solving_time_eades_four = post_time_eades_four - pre_time_eades_four
self.assertIsNotNone(output_eades_four)
score_eades_four = scoreSolution(self.matrix_input_size_1000, output_eades_four)
file_string_four = "Eades Approximation, |V| = 1000: Score = %.4f, Time = %.4f\n" % (
score_eades_four,
solving_time_eades_four,
)
self.benchmark_file.write(file_string_four)
def maximum_acyclic_subgraph_helper_efficient(self, graph):
vertices = [i for i in range(len(graph))]
shuffle(vertices)
score_vertices = scoreSolution(graph, vertices)
reverse_vertices = vertices[::-1]
score_reverse = scoreSolution(graph, reverse_vertices)
if score_vertices > score_reverse:
return vertices
else:
return reverse_vertices
def test_benchmark_two_approximation(self):
# |V| = 4
pre_time_two_approx_one = time.time()
output_two_approx_one = TwoApproximationSolver(self.matrix_input_brute_force_two).maximum_acyclic_subgraph()
post_time_two_approx_one = time.time()
solving_time_two_approx_one = post_time_two_approx_one - pre_time_two_approx_one
self.assertIsNotNone(output_two_approx_one)
score_two_approx_one = scoreSolution(self.matrix_input_brute_force_two, output_two_approx_one)
file_string_one = "Two Approximation, |V| = 4: Score = %.4f, Time = %.4f\n" % (
score_two_approx_one,
solving_time_two_approx_one,
)
self.benchmark_file.write(file_string_one)
# |V| = 25
pre_time_two_approx_two = time.time()
output_two_approx_two = TwoApproximationSolver(self.matrix_input_brute_force_five).maximum_acyclic_subgraph()
post_time_two_approx_two = time.time()
solving_time_two_approx_two = post_time_two_approx_two - pre_time_two_approx_two
self.assertIsNotNone(output_two_approx_two)
score_two_approx_two = scoreSolution(self.matrix_input_brute_force_five, output_two_approx_two)
file_string_two = "Two Approximation, |V| = 25: Score = %.4f, Time = %.4f\n" % (
score_two_approx_two,
solving_time_two_approx_two,
)
self.benchmark_file.write(file_string_two)
# |V| = 100
pre_time_two_approx_three = time.time()
output_two_approx_three = TwoApproximationSolver(self.matrix_input_size_100).maximum_acyclic_subgraph()
post_time_two_approx_three = time.time()
solving_time_two_approx_three = post_time_two_approx_three - pre_time_two_approx_three
self.assertIsNotNone(output_two_approx_three)
score_two_approx_three = scoreSolution(self.matrix_input_size_100, output_two_approx_three)
file_string_three = "Two Approximation, |V| = 100: Score = %.4f, Time = %.4f\n" % (
score_two_approx_three,
solving_time_two_approx_three,
)
self.benchmark_file.write(file_string_three)
# |V| = 1000
pre_time_two_approx_four = time.time()
output_two_approx_four = TwoApproximationSolver(self.matrix_input_size_1000).maximum_acyclic_subgraph()
post_time_two_approx_four = time.time()
solving_time_two_approx_four = post_time_two_approx_four - pre_time_two_approx_four
self.assertIsNotNone(output_two_approx_four)
score_two_approx_four = scoreSolution(self.matrix_input_size_1000, output_two_approx_four)
file_string_four = "Two Approximation, |V| = 1000: Score = %.4f, Time = %.4f\n" % (
score_two_approx_four,
solving_time_two_approx_four,
)
self.benchmark_file.write(file_string_four)
def test_benchmark_simulated_annealing(self):
# |V| = 4
pre_time_sa_one = time.time()
output_sa_one = SimulatedAnnealingSolver(
range(len(self.matrix_input_brute_force_two)), self.matrix_input_brute_force_two
).maximum_acyclic_subgraph()
post_time_sa_one = time.time()
solving_time_sa_one = post_time_sa_one - pre_time_sa_one
self.assertIsNotNone(output_sa_one)
score_sa_one = scoreSolution(self.matrix_input_brute_force_two, output_sa_one)
file_string_one = "Simulated Annealing, |V| = 4: Score = %.4f, Time = %.4f\n" % (
score_sa_one,
solving_time_sa_one,
)
self.benchmark_file.write(file_string_one)
# |V| = 25
pre_time_sa_two = time.time()
output_sa_two = SimulatedAnnealingSolver(
range(len(self.matrix_input_brute_force_five)), self.matrix_input_brute_force_five
).maximum_acyclic_subgraph()
post_time_sa_two = time.time()
solving_time_sa_two = post_time_sa_two - pre_time_sa_two
self.assertIsNotNone(output_sa_two)
score_sa_two = scoreSolution(self.matrix_input_brute_force_five, output_sa_two)
file_string_two = "Simulated Annealing, |V| = 25: Score = %.4f, Time = %.4f\n" % (
score_sa_two,
solving_time_sa_two,
)
self.benchmark_file.write(file_string_two)
# |V| = 100
pre_time_sa_three = time.time()
output_sa_three = SimulatedAnnealingSolver(
range(len(self.matrix_input_size_100)), self.matrix_input_size_100
).maximum_acyclic_subgraph()
post_time_sa_three = time.time()
solving_time_sa_three = post_time_sa_three - pre_time_sa_three
self.assertIsNotNone(output_sa_three)
score_sa_three = scoreSolution(self.matrix_input_size_100, output_sa_three)
file_string_three = "Simulated Annealing, |V| = 100: Score = %.4f, Time = %.4f\n" % (
score_sa_three,
solving_time_sa_three,
)
self.benchmark_file.write(file_string_three)
def test_benchmark_brute_force(self):
# |V| = 2
pre_time_brute_force_one = time.time()
output_brute_force_one = BruteForceSolver(self.matrix_input_brute_force_one).maximum_acyclic_subgraph()
post_time_brute_force_one = time.time()
solving_time_brute_force_one = post_time_brute_force_one - pre_time_brute_force_one
self.assertIsNotNone(output_brute_force_one)
score_brute_force_one = scoreSolution(self.matrix_input_brute_force_one, output_brute_force_one)
file_string_one = "Brute Force, |V| = 2: Score = %.4f, Time = %.4f\n" % (
score_brute_force_one,
solving_time_brute_force_one,
)
self.benchmark_file.write(file_string_one)
print ("Finished Benchmark 1, Brute Force")
# |V| = 4
pre_time_brute_force_two = time.time()
output_brute_force_two = BruteForceSolver(self.matrix_input_brute_force_two).maximum_acyclic_subgraph()
post_time_brute_force_two = time.time()
solving_time_brute_force_two = post_time_brute_force_two - pre_time_brute_force_two
self.assertIsNotNone(output_brute_force_two)
score_brute_force_two = scoreSolution(self.matrix_input_brute_force_two, output_brute_force_two)
file_string_two = "Brute Force, |V| = 4: Score = %.4f, Time = %.4f\n" % (
score_brute_force_two,
solving_time_brute_force_two,
)
self.benchmark_file.write(file_string_two)
print "Finished Benchmark 2, Brute Force"
# |V| = 8
pre_time_brute_force_three = time.time()
output_brute_force_three = BruteForceSolver(self.matrix_input_brute_force_three).maximum_acyclic_subgraph()
post_time_brute_force_three = time.time()
solving_time_brute_force_three = post_time_brute_force_three - pre_time_brute_force_three
self.assertIsNotNone(output_brute_force_three)
score_brute_force_three = scoreSolution(self.matrix_input_brute_force_three, output_brute_force_three)
file_string_three = "Brute Force, |V| = 8: Score = %.4f, Time = %.4f\n" % (
score_brute_force_three,
solving_time_brute_force_three,
)
self.benchmark_file.write(file_string_three)
print "Finished Benchmark 3, Brute Force"
def maximum_acyclic_subgraph(self):
vertices = range(len(self.adj_matrix))
vertex_permutations = permutations(vertices)
max_score = 0
max_ordering = vertices
for permutation in vertex_permutations:
score = scoreSolution(self.adj_matrix, permutation)
if score > max_score:
max_score = score
max_ordering = permutation
return list(max_ordering)
def test_benchmark_feedback_arc_set_ip(self):
# |V| = 2
pre_time_ip_one = time.time()
output_solver_one = FinalSolver(self.matrix_input_brute_force_one)
output_ip_one = output_solver_one.obtain_library_solution(output_solver_one.adj_matrix, force_ip=True)
post_time_ip_one = time.time()
solving_time_ip_one = post_time_ip_one - pre_time_ip_one
self.assertIsNotNone(output_ip_one)
score_ip_one = scoreSolution(self.matrix_input_brute_force_one, output_ip_one)
file_string_one = "Ip Approximation, |V| = 2: Score = %.4f, Time = %.4f\n" % (score_ip_one, solving_time_ip_one)
self.benchmark_file.write(file_string_one)
# |V| = 4
pre_time_ip_two = time.time()
output_solver_two = FinalSolver(self.matrix_input_brute_force_two)
output_ip_two = output_solver_two.obtain_library_solution(output_solver_two.adj_matrix, force_ip=True)
post_time_ip_two = time.time()
solving_time_ip_two = post_time_ip_two - pre_time_ip_two
self.assertIsNotNone(output_ip_two)
score_ip_two = scoreSolution(self.matrix_input_brute_force_two, output_ip_two)
file_string_two = "Ip Approximation, |V| = 4: Score = %.4f, Time = %.4f\n" % (score_ip_two, solving_time_ip_two)
self.benchmark_file.write(file_string_two)
# |V| = 8
pre_time_ip_three = time.time()
output_solver_three = FinalSolver(self.matrix_input_brute_force_three)
output_ip_three = output_solver_three.obtain_library_solution(output_solver_three.adj_matrix, force_ip=True)
post_time_ip_three = time.time()
solving_time_ip_three = post_time_ip_three - pre_time_ip_three
self.assertIsNotNone(output_ip_three)
score_ip_three = scoreSolution(self.matrix_input_brute_force_three, output_ip_three)
file_string_three = "Ip Approximation, |V| = 8: Score = %.4f, Time = %.4f\n" % (
score_ip_three,
solving_time_ip_three,
)
self.benchmark_file.write(file_string_three)
def test_final_solver(self):
'''
TODO: Add more tests, especially on our own instances.
Tests a variety of input graphs (the same as those tested for the two
approximation). Since the algorithm is randomized, the seed is set,
though this algorithm should converge towards an optimal solution at a
much faster rate, especially on graphs of this size.
Gyazo links are attached for the constructed graphs, and the first cut
made is denoted for each graph (the seed is set accordingly).
1. Test input one: https://gyazo.com/74dcc4c2eccd617381867274ce981859
2. Test input DAG one: https://gyazo.com/eae09e06bf5f4b0150977bf487399001
3. Test input DAG two: https://gyazo.com/8b241b60a8339d993fcda0b6ca6401a6
4. Test input cycle: https://gyazo.com/48c0e885e8baca8c0bd34a2f290171c8
5. Test input nonplanar one: https://gyazo.com/e91def6e27e59de97a5a779b87e659a9
6. Test input nonplanar two: https://gyazo.com/1295ad7498c8f80ff7ff260287a8d0ae
'''
random.seed(170)
self.maxDiff = None
expected_output_one = [2, 1, 3, 0]
observed_output_one = \
FinalSolver(self.matrix_input_one).maximum_acyclic_subgraph()
expected_score_one = 4.0
observed_score_one = scoreSolution(self.matrix_input_one, observed_output_one)
self.assertEquals(observed_output_one, expected_output_one)
self.assertEquals(observed_score_one, expected_score_one)
expected_output_dag_one = [0, 1, 2, 3]
observed_output_dag_one = \
FinalSolver(self.matrix_input_dag_one).maximum_acyclic_subgraph()
expected_score_dag_one = 3.0
observed_score_dag_one = scoreSolution(self.matrix_input_dag_one, observed_output_dag_one)
self.assertEquals(observed_output_dag_one, expected_output_dag_one)
self.assertEquals(observed_score_dag_one, expected_score_dag_one)
expected_output_cycle = [3, 0, 1, 2]
observed_output_cycle = \
FinalSolver(self.matrix_input_cycle).maximum_acyclic_subgraph()
expected_score_cycle = 3.0
observed_score_cycle = scoreSolution(self.matrix_input_cycle, observed_output_cycle)
self.assertEquals(observed_output_cycle, expected_output_cycle)
self.assertEquals(observed_score_cycle, expected_score_cycle)
# Could also be linearized as [1, 2, 0, 3, 4, 5, 6] but the algorithm
# should break ties arbitrarily and this is chosen
expected_output_dag_two = [1, 0, 2, 3, 4, 5, 6]
observed_output_dag_two = \
FinalSolver(self.matrix_input_dag_two).maximum_acyclic_subgraph()
expected_score_dag_one = 8.0
observed_score_dag_one = scoreSolution(self.matrix_input_dag_two, observed_output_dag_two)
self.assertEquals(observed_output_dag_two, expected_output_dag_two)
self.assertEquals(observed_score_dag_one, expected_score_dag_one)
expected_output_nonplanar_one = [8, 9, 10, 13, 0, 1, 2, 3, 4, 5, 11, 6, 7, 12]
observed_output_nonplanar_one = \
FinalSolver(self.matrix_input_nonplanar_one).maximum_acyclic_subgraph()
expected_score_nonplanar_one = 19.0
observed_score_nonplanar_one = scoreSolution(
self.matrix_input_nonplanar_one,
observed_output_nonplanar_one
)
self.assertEquals(observed_output_nonplanar_one, expected_output_nonplanar_one)
self.assertEquals(observed_score_nonplanar_one, expected_score_nonplanar_one)
expected_output_nonplanar_two = [5, 0, 7, 6, 4, 3, 1, 2]
observed_output_nonplanar_two = \
FinalSolver(self.matrix_input_nonplanar_two).maximum_acyclic_subgraph()
expected_score_nonplanar_two = 18.0
observed_score_nonplanar_two = scoreSolution(
self.matrix_input_nonplanar_two,
observed_output_nonplanar_two
)
self.assertEquals(observed_output_nonplanar_two, expected_output_nonplanar_two)
self.assertEquals(observed_score_nonplanar_two, expected_score_nonplanar_two)
observed_output_foster = \
FinalSolver(self.matrix_foster).maximum_acyclic_subgraph()
expected_score_foster = 124.0
observed_score_foster = scoreSolution(self.matrix_foster, observed_output_foster)
self.assertEquals(observed_score_foster, expected_score_foster)
observed_output_gray = \
FinalSolver(self.matrix_input_gray).maximum_acyclic_subgraph()
expected_score_gray = 75.0
observed_score_gray = scoreSolution(self.matrix_input_gray, observed_output_gray)
self.assertEquals(observed_score_gray, expected_score_gray)
observed_output_new_graph = \
FinalSolver(self.matrix_new_graph).maximum_acyclic_subgraph()
expected_score_new_graph = 55.0
observed_score_new_graph = scoreSolution(self.matrix_new_graph, observed_output_new_graph)
self.assertEquals(observed_score_new_graph, expected_score_new_graph)
def maximum_acyclic_subgraph(self):
topo_sort = DAGSolver(self.adj_matrix).topological_sort()
if topo_sort is not None:
print "Score is %.4f" % scoreSolution(self.adj_matrix, topo_sort)
return topo_sort
# For some reason, this is returned in reverse order by the algorithm.
# So reverse to get the topologically sorted SCCs.
scc_graph = self.find_strongly_connected_components()[::-1]
solution = []
for scc in scc_graph:
scc_adj_matrix = self.create_scc_adj_matrix(scc)
if len(scc_adj_matrix) <= 8:
brute_force_solver = BruteForceSolver(scc_adj_matrix)
brute_force_solution = brute_force_solver.maximum_acyclic_subgraph()
scc_solution = [i for i in range(len(scc_adj_matrix))]
for i in range(len(brute_force_solution)):
scc_conversion_index = brute_force_solution[i]
scc_solution[i] = scc[scc_conversion_index]
else:
pq = []
library_solution = self.obtain_library_solution(scc_adj_matrix)
library_score = scoreSolution(scc_adj_matrix, library_solution)
print "Finished scoring using library"
two_approx_solver = TwoApproximationSolver(scc_adj_matrix)
for i in range(1000):
this_solution = two_approx_solver.maximum_acyclic_subgraph()
this_score = scoreSolution(scc_adj_matrix, this_solution)
# If pq still small, add this solution
# Otherwise, new score better than the worst one so far
if len(pq) < 5:
heapq.heappush(pq, (this_score, this_solution))
elif this_score > pq[0][0]:
heapq.heappushpop(pq, (this_score, this_solution))
print "Finished Two Approximation"
# Add library solution last so it doesn't get overwritten
heapq.heappush(pq, (library_score, library_solution))
annealing_score = -float('inf')
annealing_solution = None
while len(pq) > 0:
curr_soln = heapq.heappop(pq)
curr_ordering = curr_soln[1]
simulated_annealing_solver = SimulatedAnnealingSolver(
curr_ordering,
scc_adj_matrix
)
curr_annealing_solution = simulated_annealing_solver.maximum_acyclic_subgraph()
curr_annealing_score = scoreSolution(scc_adj_matrix, curr_annealing_solution)
if curr_annealing_score > annealing_score:
annealing_score = curr_annealing_score
annealing_solution = curr_annealing_solution
scc_solution = [i for i in range(len(scc_adj_matrix))]
for i in range(len(annealing_solution)):
scc_conversion_index = annealing_solution[i]
scc_solution[i] = scc[scc_conversion_index]
solution.extend(scc_solution)
print "Score is %.4f" % scoreSolution(self.adj_matrix, solution)
return solution