def get_linreg_w(X, Y):
""" X: dataframe of x1, x2, x..., xn
Y: array of y
return: w as matrix """
print X
X['b'] = 1
Xt = X.transpose()
w_den = np.dot(Xt, X)
w_pre = np.dot(utils.matrix_inverse(w_den), Xt)
w = np.dot(w_pre, Y)
del X['b']
return w
def get_linridge_w(X_uncentered, Y, learning_rate):
""" Linear ridge
X: dataframe of x1, x2, x..., xn
Y: array of y
return: w as matrix """
#TODO - add mean back in before predict
X = X_uncentered
X['b'] = 1
Xt = X.transpose()
I = np.identity(X.shape[1])
w_den = np.dot(Xt, X) + np.dot(learning_rate, I)
#w_den = np.cov(X) + np.dot(learning_rate, I)
w_pre = np.dot(utils.matrix_inverse(w_den), Xt)
w = np.dot(w_pre, Y)
return w
