Gradient descent for linear regression (One variable)

GradientDescent.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
#
#  GradientDescent.py
#  
#  Copyright 2015 Overxflow

def h0(sample,theta0,theta1): return theta0+sample*theta1
def derivative_0(train_samples,theta0,theta1): return (1.0/len(train_samples))*sum([h0(sample[0],theta0,theta1)-sample[1] for sample in train_samples])
def derivative_1(train_samples,theta0,theta1): return (1.0/len(train_samples))*sum([(h0(sample[0],theta0,theta1)-sample[1])*sample[0] for sample in train_samples])

def train(train_samples,theta0,theta1,alpha,it):	
	ant_theta0,ant_theta1,i = float('inf'),float('inf'),0
	while (ant_theta0!=theta0 and ant_theta1!=theta1) or i<it:
		temp0,temp1  = theta0 - alpha*derivative_0(train_samples,theta0,theta1),theta1 - alpha*derivative_1(train_samples,theta0,theta1)
		theta0,theta1 = temp0,temp1
		ant_theta0,ant_theta1,i = theta0,theta1,i+1
	return (theta0,theta1)
	
def predict(test_sample,theta0,theta1): return h0(test_sample,theta0,theta1) 

"""if __name__ == '__main__':
	train_samples   = [(1,127),(2,325),(3,459),(4,720)]
	(theta0,theta1) = train(train_samples,900,-0.1,0.1,100000)
	print predict(2.5,theta0,theta1)
"""