def Percep(training_data, testing_data, f, coef_init, intercept_init):
ppn = Perceptron(warm_start=True)
ppn.fit(training_data, training_label)
ppn._allocate_parameter_mem(n_classes=3,
n_features=f,
coef_init=coef_init,
intercept_init=intercept_init)
ppn.fit(training_data, training_label)
weight = np.c_[(ppn.coef_), (ppn.intercept_).T]
acc_tra = ppn.score(training_data, training_label)
acc_tes = ppn.score(testing_data, testing_label)
result = (weight, acc_tra, acc_tes)
return result
# max_weight = np.concatenate((clf.intercept_.reshape(3, 1), clf.coef_), axis=1)
# print('\n')
# print("classification accuracy")
# print(max_number)
# print('\n')
# print("the final 3 weight vectors")
# print(max_weight)
# print('\n')
#13 features testing set
max_number = 0
max_weight = [[], [], []]
for i in range(100):
clf = Perceptron()
clf._allocate_parameter_mem(n_classes=3,
n_features=2,
coef_init=np.random.randn(3, 2),
intercept_init=np.random.randn(3, ))
clf.fit(train_std, l2)
acc = clf.score(test_train_std, test_l2)
if acc > max_number:
max_number = acc
max_weight = np.concatenate((clf.intercept_.reshape(3, 1), clf.coef_),
axis=1)
print('\n')
print("classification accuracy")
print(max_number)
print('\n')
print("the final 3 weight vectors")
print(max_weight)
print('\n')