def eval_classification(X, y, q=20):
X_train, X_test, y_train, y_test = testing.hold_out(X, y)
W, M = train(X_train, y_train, q=q, activation=processing.sigmoid)
D_teste = predict(X_test, W, M, activation=processing.sigmoid)
y_test = processing.encode_label(y_test)
y_pred = processing.encode_label(D_teste)
acc = round(testing.accuracy(y_test, y_pred), ndigits=2)
return acc
def eval_classification(X, y, q=Q):
X_train, X_test, y_train, y_test = testing.hold_out(X,
y,
test_size=TEST_SIZE)
T, G = train(X_train, y_train, q=q, activation=processing.sigmoid)
D_teste = predict(X_test, T, G, activation=processing.sigmoid)
y_test = processing.encode_label(y_test)
y_pred = processing.encode_label(D_teste)
acc = round(testing.accuracy(y_test, y_pred), ndigits=2)
return acc
def main():
X, y = load.iris()
X_train, X_test, y_train, y_test = testing.hold_out(X, y)
W = train(X_train, y_train)
y_pred = predict(X_test, W)
y_test = processing.encode_label(y_test)
acc = testing.accuracy(y_test, y_pred)
print("ACC: {:.2f} %".format(acc * 100))
def predict(X, W):
"""Função de predição baseado na memória W da Rede Neural Perceptron.
X é a matriz de features e não deve conter o bias, pois é
adicionado nessa função.
"""
X = processing.add_bias(X)
u = W.T @ X.T
return processing.encode_label(u.T)
def score(self, X, y):
y_pred = self.predict(X)
y_encoded = processing.encode_label(y_pred)
return testing.accuracy(y, y_encoded)
def predict(self, X):
y_pred = predict(X, self.T, self.G, activation=self.activation)
encoded = processing.encode_label(y_pred).astype(int)
return encoded