def test_nsga2_simple(self):
size = 14
master_graph = nx.grid_graph([size, size])
master_graph.graph['districts'] = 2
nx.set_node_attributes(master_graph, {i: 1
for i in master_graph},
name='pop')
nx.set_node_attributes(master_graph, {(i, j): box(i, j, i + 1, j + 1)
for (i, j) in master_graph},
name='shape')
frontier, data = genetics.run_nsga2(master_graph,
[PopulationEquality(master_graph)],
multiprocess=True,
max_generations=100,
pop_size=500,
optimize=True)
for gen_id in data:
if gen_id % 20 != 0:
continue
gen_data = data[gen_id]
random_pareto: Candidate = random.choice(
gen_data.get('pareto_frontier'))
random_pareto.plot(save=True)
random_pareto.export()
print("\n".join("{}".format(i.chromosome.get_component_scores())
for i in frontier))
def setUpClass(cls):
cls.master_graph = nx.cycle_graph(10)
cls.master_graph.graph['districts'] = 2
nx.set_node_attributes(cls.master_graph,
{i: 1
for i in cls.master_graph},
name='pop')
Chromosome.objectives = [PopulationEquality(cls.master_graph)]
def setUp(self):
self.simple_graph = nx.path_graph(4)
nx.set_node_attributes(self.simple_graph, {i: 1
for i in range(4)},
name='pop')
Chromosome.objectives = [
PopulationEquality(self.simple_graph, districts=2)
]
def test_find_best_neighbor_simple(self):
master_graph = nx.path_graph(6)
nx.set_node_attributes(master_graph, {i: 1
for i in master_graph},
name='pop')
master_graph = nx.freeze(master_graph)
Chromosome.objectives = [PopulationEquality(master_graph)]
s1 = Chromosome(master_graph, [1, 1, 2, 2, 2, 2])
s2 = find_best_neighbor(s1)
self.assertEqual(s2.get_scores(), [0.0])
self.assertEqual(s2.get_assignment(), [1, 1, 1, 2, 2, 2])
s3 = find_best_neighbor(s2)
self.assertIsNone(s3)
def test_optimize_simple(self):
master_graph = nx.path_graph(6)
nx.set_node_attributes(master_graph, {i: 1
for i in master_graph},
name='pop')
Chromosome.objectives = [PopulationEquality(master_graph)]
chromosome = Chromosome(master_graph, [1, 2, 2, 2, 2, 2])
best_state = optimize(chromosome)
self.assertEqual(best_state.get_scores(), [0.0])
self.assertEqual(best_state.get_assignment(), [1, 1, 1, 2, 2, 2])
better_state = optimize(best_state)
best_state.normalize()
better_state.normalize()
self.assertEqual(best_state, better_state)
def setUp(self):
self.master_graph = nx.Graph()
nx.add_path(self.master_graph, [0, 1, 2])
nx.add_path(self.master_graph, [5, 6, 7])
nx.add_path(self.master_graph, [2, 3, 4, 5])
self.master_graph.graph['order'] = {i: i for i in range(8)}
self.master_graph.graph['districts'] = 2
nx.set_node_attributes(self.master_graph,
{i: 1
for i in self.master_graph},
name='pop')
self.master_graph = nx.freeze(self.master_graph)
Chromosome.objectives = [PopulationEquality(self.master_graph)]
self.m_tqdm_patch = patch('elbridge.evolution.search.tqdm')
self.m_tqdm = self.m_tqdm_patch.start()
self.m_tqdm.side_effect = lambda x, *y, **z: x
def test_connect_vertices_complex_graph(self):
graph = nx.grid_graph([3, 3])
nx.set_node_attributes(graph, {i: 1 for i in graph}, name='pop')
Chromosome.objectives = [PopulationEquality(graph, districts=2)]
chromosome = Chromosome(graph, [1, 1, 1, 1, 1, 1, 2, 2, 2])
chromosome = chromosome.connect_vertices(((1, 1), (2, 1)))
self.assertEqual(chromosome.get_assignment(),
[1, 1, 1, 1, 1, 1, 2, 1, 2])
chromosome = chromosome.connect_vertices(((2, 1), (2, 0)))
self.assertEqual(chromosome.get_assignment(),
[1, 1, 1, 1, 1, 1, 1, 1, 2])
chromosome = chromosome.connect_vertices(((2, 1), (2, 2)))
self.assertEqual(chromosome.get_assignment(),
[1, 1, 1, 1, 1, 1, 1, 1, 1])
def test_optimize_moderately_complex(self):
grid_size = [32, 32]
node_count = grid_size[0] * grid_size[1]
edge_count = (grid_size[0] -
1) * grid_size[1] + grid_size[0] * (grid_size[1] - 1)
districts = 2
assignment = random.choices(range(1, districts + 1), k=node_count)
master_graph = nx.grid_graph(grid_size)
nx.set_node_attributes(master_graph, {i: 1
for i in master_graph},
name='pop')
master_graph.graph['districts'] = districts
Chromosome.objectives = [PopulationEquality(master_graph)]
chromosome = Chromosome(master_graph, assignment)
best_state = optimize(chromosome, sample_size=edge_count * 2)
self.assertTrue(best_state.dominates(chromosome))
def test_find_best_neighbor_complex(self):
size = 1000
master_graph = nx.path_graph(size)
nx.set_node_attributes(master_graph, {i: 1
for i in range(2)},
name='pop')
nx.set_node_attributes(master_graph, {i: 2
for i in range(2, size)},
name='pop')
master_graph.graph['districts'] = 2
Chromosome.objectives = [PopulationEquality(master_graph)]
chromosome = Chromosome(master_graph, list(range(size)))
new_chromosome = find_best_neighbor(chromosome, sample_size=size * 2)
self.assertEqual(new_chromosome.get_scores(), [0])
self.assertEqual(new_chromosome.get_assignment(),
[1, 1] + list(range(2, size)))