def _has_data_status(self):
self.has_data = True
data = LabeledDatasetFromFile(self.data_file_path).get_data_by_sheet(0)
self.raw_data = data
self.raw_data[self.raw_data[:, -1] == -1,
-1] = 0. # default for 0/1 binary classification problem
# more classes is also ok, e.g. 4 classes is 0,1,2,3
self.output_nodes_spin_box.setEnabled(True)
self.layer_pos_spin_box.setEnabled(True)
self.insert_layer_btn.setEnabled(True)
self.remove_layer_btn.setEnabled(True)
self.node_pos_spin_box.setEnabled(True)
self.insert_node_btn.setEnabled(True)
self.remove_node_btn.setEnabled(True)
self.ratio_slider.setEnabled(True)
self.learning_rate_line_edit.setEnabled(True)
self.max_epoch_line_edit.setEnabled(True)
self.train_btn.setEnabled(True)
self.save_btn.setEnabled(True)
self.load_btn.setEnabled(True)
self.label7.setText("Now you can train your train data.")
plt.plot(x, y, '*r')
else:
plt.plot(x, y, '*g')
plt.title(title)
plt.show()
end_time = time.time()
print('cost time for visualization (title: %s) is %d sec.' %
(title, end_time - start_time))
#%%
if __name__ == '__main__':
# 调整数据格式
data_address = r'..\dataset\demodata.xls'
datasetff = LabeledDatasetFromFile(data_address).get_data_by_sheet(0)
datasetff.astype(np.float)
datasetff = datasetff[:, -3:] # 只使用连续属性值
datasetff[datasetff[:, -1] == 0, -1] = -1
train_data = datasetff[:100, :]
test_data = datasetff[100:, :]
dataset = LabeledTrainAndTestDataset(train_data, test_data=test_data)
# dataset.visual_data(train_data)
# 训练模型
svm = SupportVectorMachine(train_data,
test_data,
epsilon=0.0001,
C=200,
kernel_option=('rbf', 1.3))
turtle.goto(end_x, end_y)
turtle.penup()
turtle.goto((start_x + end_x) / 2., (start_y + end_y) / 2.)
turtle.pendown()
turtle.write('final model')
turtle.mainloop()
if __name__ == '__main__':
def f(x):
return 1 if x > 0 else 0
data_address = r'..\dataset\watermelon3.xlsx'
datasetff = LabeledDatasetFromFile(data_address).get_data_by_sheet(
0, mode=CONSTANT.TRANS)
datasetff.astype(np.float)
train_data = datasetff[:, -3:] # 只使用连续属性值
linear_separable_data = train_data[
[0, 1, 2, 3, 4, 6, 8, 9, 10, 11, 12, 13, 16], :]
classifier = Perceptron(linear_separable_data,
linear_separable_data,
f,
epsilon=0.000001)
classifier.train(max_epoch=1000, learning_rate=0.00001)
# classifier.visual_train_data_and_model(visual_process=True, step=200)
classifier.visualize_train_data_and_model_with_turtle()
error = classifier.evaluate_train_data()
if error != -1:
print('error rate: %f' % error)
projected_y = projected_x * k
plt.plot(projected_x, projected_y, '*g')
plt.plot([x[0], projected_x], [x[1], projected_y],
color="green",
linestyle=":")
plt.title('Linear Discrimenant Analysis')
plt.xlabel('密度')
plt.ylabel('含糖率')
plt.show()
#%%
if __name__ == '__main__':
data_address = r'..\dataset\watermelon3.xlsx'
datasetff = LabeledDatasetFromFile(data_address)
entire_train_data = datasetff.get_data_by_sheet(
0, mode=MyMachineLearning.utils.CONSTANT.TRANS)
entire_train_data.astype(np.float)
selected_train_data = datasetff.get_data_by_sheet(
0,
mode=MyMachineLearning.utils.CONSTANT.TRANS,
feat_indces=[6, 7],
label_indces=[8])
selected_train_data.astype(np.float)
# 线性回归
print('1. Linear Regression:')
regression_data_address = r'D:\Project\Github\LearningMachineLearning\dataset\salary_data.csv'
regrsssion_data = LabeledDatasetFromFile(
regression_data_address,
def test(self):
if not self._is_trained:
return
x = self.test_data[:, :2]
y = self.test_data[:, 2]
out = self.net(x)
pred = torch.max(out, 1)[1]
pred_y = pred.data.numpy()
target_y = y.data.numpy()
accuracy = float(
(pred_y == target_y).astype(int).sum()) / float(target_y.size)
return accuracy
if __name__ == '__main__':
data_address = r'..\..\dataset\demodata.xls'
train_data = LabeledDatasetFromFile(data_address).get_data_by_sheet(0)
train_data[train_data[:, 2] == -1, 2] = 0. # preprocess
train_data.astype(np.float)
np.random.shuffle(train_data)
test_data = torch.from_numpy(train_data[150:, :])
train_data = torch.from_numpy(train_data[:150, :])
fcnn = SimpleFCNN([2, 7, 5, 2], train_data, test_data)
fcnn.train(max_epoch=20000)
print(fcnn.test())