def test_fixed(self):
G1 = nx.MultiDiGraph()
G1.add_edge(0, 0, action_number=0)
G1.add_edge(0, 1, action_number=1)
G1.add_edge(1, 2, action_number=0)
G1.add_edge(2, 1, action_number=0)
G1.add_edge(2, 0, action_number=1)
G1.add_edge(2, 5, action_number=2)
for node in G1:
G1.node[node]['type_id'] = node
G1.node[node]['parameters'] = pyevolve.G1DList.G1DList(0)
G2 = nx.MultiDiGraph()
G2.add_edge(0, 1, action_number=0)
G2.add_edge(0, 3, action_number=1)
G2.add_edge(3, 4, action_number=0)
G2.add_edge(1, 4, action_number=0)
for node in G2:
G2.node[node]['type_id'] = node
G2.node[node]['parameters'] = pyevolve.G1DList.G1DList(0)
nodes_degrees = [2, 1, 3, 1, 0, 0]
nodes_params = [0] * len(nodes_degrees)
genome1 = graph_genome.GraphGenome(len(G1),
nodes_degrees,
nodes_params
)
genome1.graph = G1
genome2 = graph_genome.GraphGenome(len(G2),
nodes_degrees,
nodes_params
)
genome2.graph = G2
genome1.prune_non_connected()
self.assertEqual(len(genome1), 4)
self.assertEqual(len(genome2), 4)
sanity_checks(self, genome1)
sanity_checks(self, genome2)
bogus = str(genome1)
bogus = str(genome2)
genome1, genome2 = graph_genome.graph_crossover(None,
mom=genome1,
dad=genome2)
sanity_checks(self, genome1)
sanity_checks(self, genome2)
def test_pickling(self):
num_nodes = 100
genome = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params)
genome.initialize()
state = cPickle.dumps(genome)
newgenome = cPickle.loads(state)
self.assertEqual(len(genome), len(newgenome))
def test_iterator(self):
num_nodes = 100
genome = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params)
genome.initialize()
all_nodes = [node for node in genome.nodes]
self.assertEqual(len(all_nodes), len(genome))
sanity_checks(self, genome)
def test_mutation_crossover(self):
num_nodes_dad = random.randint(5,100)
num_nodes_mom = random.randint(5,100)
num_nodes_types = random.randint(1, 100)
nodes_degrees = [random.randint(0,50) for _ in xrange(num_nodes_types)]
nodes_params = [random.randint(0,50) for _ in xrange(num_nodes_types)]
dad = graph_genome.GraphGenome(num_nodes_dad,
nodes_degrees,
nodes_params
)
mom = graph_genome.GraphGenome(num_nodes_mom,
nodes_degrees,
nodes_params
)
mom.setParams(p_del=0.5, p_add=0.5)
dad.setParams(p_del=0.5, p_add=0.5)
mom.initialize()
dad.initialize()
sanity_checks(self, mom)
sanity_checks(self, dad)
for step in xrange(300):
sister, brother = graph_genome.graph_crossover(None, mom=mom, dad=dad)
sanity_checks(self, sister)
sanity_checks(self, brother)
sanity_check_all_connected(self, sister)
sanity_check_all_connected(self, brother)
sister.mutate(pmut=0.5, ga_engine = None)
brother.mutate(pmut=0.5, ga_engine = None)
sanity_checks(self, sister)
sanity_checks(self, brother)
sanity_check_all_connected(self, sister)
sanity_check_all_connected(self, brother)
dad = brother
mom = sister
def test_puning(self):
for _ in xrange(100):
num_nodes = 100
genome = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params)
genome.initialize()
genome.prune_non_connected()
sanity_checks(self, genome)
sanity_check_all_connected(self, genome)
def test_crossover(self):
num_nodes = 15
dad = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params
)
mom = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params
)
mom.setParams(p_del=0.0, p_add=0.0)
dad.setParams(p_del=0.0, p_add=0.0)
mom.initialize()
dad.initialize()
sanity_checks(self, mom)
sanity_checks(self, dad)
for step in xrange(100):
mom, dad = graph_genome.graph_crossover(None, mom=mom, dad=dad)
sanity_checks(self, mom)
sanity_checks(self, dad)
def mutation_test(self):
num_nodes = 20
genome = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params)
genome.initialize()
sanity_checks(self, genome)
genome.setParams(p_del=0.5, p_add=0.5)
for _ in xrange(100):
genome.mutate(pmut=0.2)
sanity_checks(self, genome)
sanity_check_all_connected(self, genome)
def test_single_add_node(self):
num_nodes = 10
genome = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params)
genome.initialize()
length = len(genome)
sanity_checks(self, genome)
genome.add_random_node()
self.assertEqual(len(genome), length + 1)
sanity_checks(self, genome)
def test_single_remove_node(self):
num_nodes = 20
genome = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params)
genome.initialize()
sanity_checks(self, genome)
genome.setParams(p_del=0.5, p_add=0.5)
length = len(genome)
genome.remove_random_node()
self.assertEqual(len(genome), length - 1)
sanity_checks(self, genome)
def test_adding_removing(self):
num_nodes = 20
genome = graph_genome.GraphGenome(num_nodes,
self.nodes_degrees,
self.nodes_params)
genome.initialize()
sanity_checks(self, genome)
genome.setParams(p_del=0.5, p_add=0.5)
for _ in xrange(100):
length = len(genome)
genome.remove_random_node()
self.assertEqual(len(genome), length - 1)
sanity_checks(self, genome)
length = len(genome)
genome.add_random_node()
self.assertEqual(len(genome), length + 1)
sanity_checks(self, genome)
return c_sum * p_sum
def eval_func(chromosome, **args):
G = chromosome.graph
return nx.average_clustering(nx.Graph(G))
def step_checks(ga_engine):
for genome in ga_engine.getPopulation():
sanity_checks(genome)
out_degrees = [1, 2, 3]
node_params = [0, 0, 0]
genome = graph_genome.GraphGenome(25, out_degrees, node_params)
genome.evaluator.set(eval_func2)
genome.setParams(p_del=0.5, p_add=0.5)
ga = GSimpleGA.GSimpleGA(genome)
ga.setElitism(True)
ga.selector.set(Selectors.GRouletteWheel)
ga.getPopulation().setParams(tournamentPool=500)
ga.setGenerations(10000)
ga.setPopulationSize(500)
ga.setCrossoverRate(0.78)
ga.setMutationRate(0.08)
#ga.setMinimax(Consts.minimaxType["maximize"])
ga.setMinimax(Consts.minimaxType["minimize"])