In [1]:

    

import numpy as np


    

In [2]:

    

def linearRegCostFunction(theta, X, y, Lambda):
    '''

        theta = conjunto de pesos
        X = características (features)
        y = alvo (target)
        Lambda = Taxa de regularização
    '''
    
    m = y.size
    H = X.dot(theta)
    J = (1/(2*m)) * np.sum(np.square(H-y)) 
    Reg = (Lambda/(2*m)) * np.sum(np.square(theta))
   
    return J+Reg


    

In [3]:

    

def linearRegGradient(theta, X, y, Lambda):
    '''

        theta = conjunto de pesos
        X = características (features)
        y = alvo (target)
        Lambda = Taxa de regularização
    '''
    
    m = y.size
    h = X.dot(theta.reshape(-1,1))
    Reg = (Lambda/m)*np.r_[[[0]],theta[1:].reshape(-1,1)]
    grad = (1/m)*(X.T.dot(h-y))+ Reg 
        
    return grad.flatten()