Skip to content

Instantly share code, notes, and snippets.

Created October 28, 2011 03:57
Show Gist options
  • Save marcelcaraciolo/1321585 to your computer and use it in GitHub Desktop.
Save marcelcaraciolo/1321585 to your computer and use it in GitHub Desktop.
multivariate linear regression
from numpy import loadtxt, zeros, ones, array, linspace, logspace, mean, std, arange
from mpl_toolkits.mplot3d import Axes3D
import matplotlib.pyplot as plt
from pylab import plot, show, xlabel, ylabel
#Evaluate the linear regression
def feature_normalize(X):
Returns a normalized version of X where
the mean value of each feature is 0 and the standard deviation
is 1. This is often a good preprocessing step to do when
working with learning algorithms.
mean_r = []
std_r = []
X_norm = X
n_c = X.shape[1]
for i in range(n_c):
m = mean(X[:, i])
s = std(X[:, i])
X_norm[:, i] = (X_norm[:, i] - m) / s
return X_norm, mean_r, std_r
def compute_cost(X, y, theta):
Comput cost for linear regression
#Number of training samples
m = y.size
predictions =
sqErrors = (predictions - y)
J = (1.0 / (2 * m)) *
return J
def gradient_descent(X, y, theta, alpha, num_iters):
Performs gradient descent to learn theta
by taking num_items gradient steps with learning
rate alpha
m = y.size
J_history = zeros(shape=(num_iters, 1))
for i in range(num_iters):
predictions =
theta_size = theta.size
for it in range(theta_size):
temp = X[:, it]
temp.shape = (m, 1)
errors_x1 = (predictions - y) * temp
theta[it][0] = theta[it][0] - alpha * (1.0 / m) * errors_x1.sum()
J_history[i, 0] = compute_cost(X, y, theta)
return theta, J_history
#Load the dataset
data = loadtxt('ex1data2.txt', delimiter=',')
#Plot the data
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
n = 100
for c, m, zl, zh in [('r', 'o', -50, -25)]:
xs = data[:, 0]
ys = data[:, 1]
zs = data[:, 2]
ax.scatter(xs, ys, zs, c=c, marker=m)
ax.set_xlabel('Size of the House')
ax.set_ylabel('Number of Bedrooms')
ax.set_zlabel('Price of the House')
X = data[:, :2]
y = data[:, 2]
#number of training samples
m = y.size
y.shape = (m, 1)
#Scale features and set them to zero mean
x, mean_r, std_r = feature_normalize(X)
#Add a column of ones to X (interception data)
it = ones(shape=(m, 3))
it[:, 1:3] = x
#Some gradient descent settings
iterations = 100
alpha = 0.01
#Init Theta and Run Gradient Descent
theta = zeros(shape=(3, 1))
theta, J_history = gradient_descent(it, y, theta, alpha, iterations)
print theta, J_history
plot(arange(iterations), J_history)
ylabel('Cost Function')
#Predict price of a 1650 sq-ft 3 br house
price = array([1.0, ((1650.0 - mean_r[0]) / std_r[0]), ((3 - mean_r[1]) / std_r[1])]).dot(theta)
print 'Predicted price of a 1650 sq-ft, 3 br house: %f' % (price)
Copy link

you algoritm gives wrong predict price, try to use test on training sets, and you will see error
you put incorrect
iterations = 100
alpha = 0.01

such quantity not enough with such alpha to find local optimum ,
for example these better
iterations = 10000
alpha = 0.001

Copy link

redcho commented Apr 2, 2016

You should flatten your predictions in the compute_cost function. Otherwise it returns array.

Copy link

I don't get the point of obtaining the size by m=y.size, and then set y.shape=(m,1).
Does anybody have any idea on what the latter does?

Copy link

Need this gradient please help!!!!


Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment