def __addLayer_convex(self, mutation):
graph_dna = Funct.DNA2graph(self.center_key)
parent_layers = []
acum_parents = 0
acum_history = []
for layer_index in range(self.current_num_layer):
node = graph_dna.key2node.get(layer_index)
parents = len(node.parents)
parent_layers.append(parents)
acum_parents += parents
acum_history.append(acum_parents)
factor = random.randint(0, sum(parent_layers))
selected_index = -1
for i in range(len(acum_history)):
if factor <= acum_history[i]:
selected_index = i
break
direction_accepted = self.verify_direction(layer=selected_index, mutation=self.mutations[mutation])
if direction_accepted == True:
self.predicted_actions.append( (selected_index, self.mutations[mutation]) )
def no_con_last_layer(DNA, max=None):
if DNA:
g = Fun.DNA2graph(DNA)
node = g.key2node.get(Fun.num_layers(DNA) - 1)
if len(node.parents) == 1:
return DNA
else:
return False
def no_con_image(DNA, max=None):
if DNA:
g = Fun.DNA2graph(DNA)
node = g.key2node.get(-1)
if len(node.kids) == 1:
return DNA
else:
return False
def con_image(DNA, max=None):
if DNA:
g = Fun.DNA2graph(DNA)
condition = all(
[node_con_image(g, node) for node in list(g.key2node.values())])
if condition:
return DNA
else:
return None
def max_parents(DNA, max):
if DNA:
g = Fun.DNA2graph(DNA)
if all([
len(node.parents) < max + 1
for node in list(g.key2node.values())
]):
return DNA
else:
return False
def is_convex(DNA, index):
g = Fun.DNA2graph(DNA)
node = g.key2node.get(index)
output = node.objects[0][1] == max(
[node.objects[0][2] for node in node.parents])
return output