def __add_rbf_neuron(rbf, pattern, label, theta_minus):
rbf_neuron = RBFNeuron(centroid=pattern, label=label)
rbf_neuron.set_weight(1.0)
max_radius = DDA.__max_radius(rbf, rbf_neuron.get_centroid(), label, theta_minus)
rbf_neuron.set_radius(max_radius)
rbf.get_rbf_neurons().append(rbf_neuron)
def train(data, target, rbf, theta_plus=.4, theta_minus=.1, epochs=1):
#criando primeiro neuronio rbf e o adicionando na rede
first_neuron = RBFNeuron(centroid=data[0], label=target[0], radius=numpy.inf)
first_neuron.set_weight(1.0)
rbf.get_rbf_neurons().append(first_neuron)
itered = 0
# executando a quantidade de epocas desejada
while itered < epochs:
itered += 1
print itered
# para cada padrão...
for pattern, label in zip(data[1:len(data)], target[1:len(target)]):
label_rbfneurons = rbf.get_rbf_for_label(label)
if len(label_rbfneurons) > 0 :
exist_neuron = False
#só os neuronios do label
for rbf_neuron in label_rbfneurons:
if rbf_neuron.output(pattern) >= theta_plus:
exist_neuron = True
rbf_neuron.set_weight(rbf_neuron.get_weight() + 1.0)
break;
if not exist_neuron:
DDA.__add_rbf_neuron(rbf, pattern, label, theta_minus)
else:
DDA.__add_rbf_neuron(rbf, pattern, label, theta_minus)
max_radius = DDA.__max_radius(rbf=rbf, rbf_centroid=pattern, label=label, theta_minus=theta_minus)
#atribuindo o maior raio a todos os neuronios que não forem da classe label
for other_label_rbf_neuron in rbf.get_rbf_for_not_label(label):
other_label_rbf_neuron.set_radius(max_radius)
