def main():
X_normalize, Y_normalize, Origin_X_normalize, Y_mean, Y_std, Origin_X_mean, Origin_X_std = load_data.load_and_normalize(
'data2.txt')
alpha = 0.024
iterations = 800
theta, j_list = gradient_descent(X_normalize, Y_normalize, alpha,
iterations)
print('The linear regression formula is {0} + {1} * x1 + {2} * x2'.format(
theta[0, 0], theta[1, 0], theta[2, 0]))
iteration_array = range(iterations)
plot_data.plot(iteration_array, j_list, 'iteration', "lost", {
'fmt': 'b-',
'title': 'Lost',
'show': True
})
house_size = 1650
br_num = 3
house_size_normal = (house_size - Origin_X_mean[0]) / Origin_X_std[0]
br_num = (br_num - Origin_X_mean[1]) / Origin_X_std[1]
X_new_normalize = np.array([1, house_size_normal, br_num])
price_normalize = X_new_normalize.dot(theta).sum()
price = price_normalize * Y_std[0] + Y_mean[0]
print('price is {0}'.format(price))
def plot(X, show = False):
plot_data.plot(X[:, 0], X[:, 1], 'x1', 'x2',
{
'markersize': 5,
'fmt': 'kx',
'show': show
})
def plot(X):
plot_data.plot(X[:, 0], X[:, 1], 'x1', 'x2',
{
'markersize': 4,
'fmt': 'bx',
'show': True
})
def main():
data = load_data.load('data1.txt')
X = data[:, 0]
Y = data[:, 1]
plot_data.plot(X, Y, 'house size', 'house price', {
'show': True,
'title': 'Original Data',
'fmt': 'rx'
})
def plot(X, idx_arr, centroids):
colors = ['r', 'y', 'g', 'b']
shapes = ['x', 'x', 'x', 'x']
cluster_count = centroids.shape[0]
for idx in range(cluster_count):
row_idx_arr = np.argwhere(idx_arr == idx).ravel()
X_idx = X[row_idx_arr]
plot_data.plot(X_idx[:, 0], X_idx[:, 1], 'x1', 'x2', {
'fmt': colors[idx] + shapes[idx],
'show': False
})
plot_data.plot(centroids[idx, 0], centroids[idx, 1], 'x1', 'x2', {
'fmt': 'k+',
'show': idx == cluster_count - 1
})
def main():
data = load_data.load('data1.txt')
pos_values = data[(data[:, 2] == 1)]
neg_values = data[(data[:, 2] == 0)]
plot_data.plot(pos_values[:, 0], pos_values[:, 1], 'score1', 'score2', {
'fmt': 'bx',
'markersize': 5
})
plot_data.plot(neg_values[:, 0], neg_values[:, 1], 'score1', 'score2', {
'fmt': 'yo',
'markersize': 5,
'show': True
})
def main():
X, y, origin_X = load_data.load_and_process('data1.txt')
initial_theta = np.zeros(X.shape[1])
res = minimize(cost_function, initial_theta, args=(X, y), method=None, jac=gradient, options={'maxiter': 400})
print(res)
theta = res.x
data = load_data.load('data1.txt')
pos_values = data[(data[:, 2] == 1)]
neg_values = data[(data[:, 2] == 0)]
plot_data.plot(pos_values[:, 0], pos_values[:, 1], 'score1', 'score2',
{
'fmt': 'bx',
'markersize': 5
})
plot_data.plot(neg_values[:, 0], neg_values[:, 1], 'score1', 'score2',
{
'fmt': 'yo',
'markersize': 5,
'show': False
})
score1 = np.linspace(25, 100)
score2 = []
for item in score1:
score2.append(((0.5 - theta[0]) - theta[1] * item) / theta[2])
score2 = np.array(score2)
plot_data.plot(score1, score2, 'score1', 'score2',
{
'fmt': 'r-',
'label': 'minimize',
'show': False
})
theta_from_cal = np.array([-4.81180027, 0.04528064, 0.03819149]) # 0.001 10 0000
theta_from_cal_2 = np.array([-15.39517866, 0.12825989, 0.12247929]) # 0.001 100 0000
theta_from_cal_3 = np.array([-22.21628108, 0.18268725, 0.17763448]) # 0.003 100 0000
score3 = []
for item in score1:
score3.append(((0.5 - theta_from_cal_3[0]) - theta_from_cal_3[1] * item) / theta_from_cal_3[2])
score3 = np.array(score3)
plot_data.plot(score1, score3, 'score1', 'score2',
{
'fmt': 'g-',
'label': 'gredient-descent',
'legend_loc': 'upper right',
'show': True
})
print(predict(45, 85, theta_from_cal_3))
def main():
x = np.linspace(-8, 8, 1000)
y = sigmoid.sigmoid(x)
plot_data.plot(x, y, 'x', 'y', {
'fmt': 'b-',
'title': 'sigmoid',
'label': 'sigmoid',
'show': False
})
y2 = []
for _ in x:
y2.append(0.5)
y2 = np.array(y2)
plot_data.plot(x, y2, 'x', 'y', {
'fmt': 'g-',
'label': 'y = 0.5',
'show': True
})
def main():
X, Y, Origin_X = load_data.load_and_process('data1.txt')
alpha = 0.023
iterations = 400
theta, j_list = gradient_descent(X, Y, alpha, iterations)
print(theta)
plot_data.plot(Origin_X, Y, 'original data figure', 'house size', {
'fmt': 'rx',
'label': 'Original Data'
})
plot_data.plot(
Origin_X, X.dot(theta), "house size", 'price', {
'fmt': 'b-',
'title': 'gradient decent',
'label': 'Linear Regression',
'legend_loc': 'lower right',
'show': True
})
iteration_array = range(iterations)
plot_data.plot(iteration_array, j_list, 'iteration', "lost", {
'fmt': 'b-',
'title': 'Lost',
'show': True
})
def main():
x, y, origin_x = load_data.load_and_process('data1.txt')
alpha = 0.003
iterations = 1000000
theta, j_list = gradient_descent(x, y, alpha, iterations)
print(theta)
iteration_array = range(iterations)
plot_data.plot(iteration_array, j_list, 'iteration', "lost", {
'fmt': 'b-',
'title': 'Lost',
'show': True
})
data = load_data.load('data1.txt')
pos_values = data[(data[:, 2] == 1)]
neg_values = data[(data[:, 2] == 0)]
plot_data.plot(pos_values[:, 0], pos_values[:, 1], 'score1', 'score2', {
'fmt': 'bx',
'markersize': 5
})
plot_data.plot(neg_values[:, 0], neg_values[:, 1], 'score1', 'score2', {
'fmt': 'yo',
'markersize': 5,
'show': False
})
score1 = np.linspace(25, 100)
score2 = []
for item in score1:
score2.append(((0.5 - theta[0, 0]) - theta[1, 0] * item) / theta[2, 0])
score2 = np.array(score2)
plot_data.plot(score1, score2, 'score1', 'score2', {
'fmt': 'r-',
'show': True
})
def plot(X, show=False):
plot_data.plot(X[:, 0], X[:, 1], 'x1', 'x2', {'show': show})