def cost_vs_iterations_plotting(learning_rates_list):
start = time.time()
if not learning_rates_list:
learning_rates_list = [.1, .5, 1]
# Dataset list
datasets = [
Datasets.IONOSPHERE, Datasets.ADULT, Datasets.WINE_QUALITY,
Datasets.BREAST_CANCER_DIAGNOSIS
]
# Initialize model
classifier = LogisticRegression()
for dataset_name in datasets:
print("dataset: ", dataset_name)
# Load the datasets
X, y = get_dataset(dataset_name)
# Feature Scalling
X = feature_scaling(X)
# Split the datasets
X_train, X_test, y_train, y_test = train_test_split(X,
y,
0.8,
shuffle=True)
for lr in learning_rates_list:
classifier.lr = lr
# Fit the model to the dataset
classifier.fit(X_train, y_train)
# Plot the evolution of the cost during training
plt.plot(range(len(classifier.cost_history)),
classifier.cost_history)
legends = []
for l in learning_rates_list:
l_str = 'lr = ' + str(l)
legends.append(l_str)
plt.legend(legends, loc='upper right')
plt.title(dataset_name)
plt.xlim((0, 100))
plt.ylabel('Cost')
plt.xlabel('Iterations')
plt.show()
print('\n\nDONE!')
print('It took', time.time() - start, 'seconds.')
assert dW.shape == self.W.shape
self.W -= self.lr * dW
self.b -= self.lr * db
def predict(self, X):
# 将矩阵压缩成向量，与原始输入Y保持一致
return np.squeeze(np.dot(X, self.W) + self.b)
def RMSE(y_true, y_pred):
return sum((y_true - y_pred) ** 2) ** 0.5 / len(y_true)
if __name__ == "__main__":
from datasets.dataset import load_boston
from model_selection.train_test_split import train_test_split
data = load_boston()
X = data.data
Y = data.target
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.2)
line_reg = LinearRegression(max_iter=2000)
line_reg.fit(X_train, Y_train)
Y_pred = line_reg.predict(X_test)
rmse = RMSE(Y_test, Y_pred)
print(rmse)
#### E-step，计算概率 ####
return np.argmax(P_mat, axis=1)
if __name__ == '__main__':
from sklearn.datasets.samples_generator import make_blobs
from model_selection.train_test_split import train_test_split
X, _ = make_blobs(cluster_std=1.5,
random_state=42,
n_samples=1000,
centers=3)
X = np.dot(X, np.random.RandomState(0).randn(2, 2)) # 生成斜形类簇
import matplotlib.pyplot as plt
plt.clf()
plt.scatter(X[:, 0], X[:, 1], alpha=0.3)
plt.show()
X_train, X_test = train_test_split(X, test_size=0.2)
n_samples, n_feature = X_train.shape
gmm = GaussianMixture(n_components=6)
gmm.fit(X_train)
Y_pred = gmm.predict(X_test)
plt.clf()
plt.scatter(X_test[:, 0], X_test[:, 1], c=Y_pred, alpha=0.3)
plt.show()
clus_pred = np.argmin(dist_test, axis=1)
return clus_pred
if __name__ == '__main__':
import numpy as np
data_1 = np.random.randn(200, 2) + [1, 1]
data_2 = np.random.randn(200, 2) + [4, 4]
data_3 = np.random.randn(200, 2) + [7, 1]
data = np.concatenate((data_1, data_2, data_3), axis=0)
from model_selection.train_test_split import train_test_split
X_train, X_test = train_test_split(data, test_size=0.2)
kmeans = KMeans(n_clusters=3)
kmeans.fit(X_train)
import matplotlib.pyplot as plt
plt.clf()
plt.scatter(X_train[:, 0], X_train[:, 1], alpha=0.5, c=kmeans.labels_)
plt.scatter(kmeans.cluster_centers_[:, 0],
kmeans.cluster_centers_[:, 1],
marker='*',
c='k')
plt.show()
clus_pred = kmeans.predict(X_test)
# Feature scaling
X = feature_scaling(X)
# Create the classifiers
lr_classifier = LogisticRegression()
nb_classifier = GaussianNaiveBayes()
train_sizes = np.arange(0.05, 1, 0.05)
lr_accuracy = []
nb_accuracy = []
for t in train_sizes:
# Split into train and test
X_train, X_test, y_train, y_test = train_test_split(X,
y,
t,
shuffle=True)
# Train and evaluate the models
lr_classifier.fit(X_train, y_train)
y_pred = lr_classifier.predict(X_test)
lr_accuracy.append(evaluate_acc(y_test, y_pred))
nb_classifier.fit(X_train, y_train)
y_pred = nb_classifier.predict(X_test)
nb_accuracy.append(evaluate_acc(y_test, y_pred))
printAccuracyComparison(ds, lr_accuracy, nb_accuracy, train_sizes)
print('\n\nDONE!')
