def de_mean(A):
""" returns result of subtracting from every value of A the mean value
of its column """
num_rows, num_cols = Ch4.shape(A)
col_means, _ = scale(A)
return Ch4.make_matrix(num_rows,num_cols,
lambda i,j: A[i][j]-col_means[j])
def de_mean(A):
""" returns result of subtracting from every value of A the mean value
of its column """
num_rows, num_cols = Ch4.shape(A)
col_means, _ = scale(A)
return Ch4.make_matrix(num_rows, num_cols,
lambda i, j: A[i][j] - col_means[j])
plt.ylabel('actual')
# Compute beta_hat by SGD #
###########################
# pick a random initial beta (constant, beta1*experience, beta2*salary)
beta_0 = [random.random() for _ in range(3)]
beta_hat = maximize_stochastic(logistic_log_likelihood_i,
logistic_log_gradient_i, x_train, y_train,
beta_0)
print "beta_hat", beta_hat
# Transform beta_hat back to unscaled variables #
#################################################
# get the means and stds of const, years, experience cols in data
means_x, stds_x = scale(x)
# beta_i i!=0 has the following transform beta_i = beta_i_scaled/sigma_i
# and beta_0 is
beta_hat_unscaled = [
beta_hat[0], beta_hat[1] / stds_x[1], beta_hat[2] / stds_x[2]
]
print "beta_hat_unscaled", beta_hat_unscaled
# Fit Quality #
###############
# Examine the test data
true_positives = false_positives = true_negatives = false_negatives = 0
for x_i, y_i in zip(x_test, y_test):
# For the test data get a prediction for y. This will be a
plt.ylabel('actual')
# Compute beta_hat by SGD #
###########################
# pick a random initial beta (constant, beta1*experience, beta2*salary)
beta_0 = [random.random() for _ in range(3)]
beta_hat = maximize_stochastic(logistic_log_likelihood_i,
logistic_log_gradient_i,
x_train, y_train, beta_0)
print "beta_hat", beta_hat
# Transform beta_hat back to unscaled variables #
#################################################
# get the means and stds of const, years, experience cols in data
means_x, stds_x = scale(x)
# beta_i i!=0 has the following transform beta_i = beta_i_scaled/sigma_i
# and beta_0 is
beta_hat_unscaled =[beta_hat[0],
beta_hat[1]/stds_x[1],
beta_hat[2]/stds_x[2]]
print "beta_hat_unscaled", beta_hat_unscaled
# Fit Quality #
###############
# Examine the test data
true_positives = false_positives = true_negatives = false_negatives = 0
for x_i, y_i in zip(x_test, y_test):
# For the test data get a prediction for y. This will be a
