def cross_validation(t, X, J, seed, lamb, istraining):
# concate t and X and named it X' and go to next step
X_prime = np.column_stack((t, X))
X_prime = X_prime.tolist()
trainning = X_prime
# random split X' to J fold -> A list of list of list
total_length = len(trainning)
fold = int(total_length / J)
random.Random(seed).shuffle(trainning)
generator = (trainning[i:i + fold] for i in range(0, len(trainning), fold))
master_fold_list = list(generator)
MSE_list = list()
# loop on index of each fold:
for v_index in range(len(master_fold_list)):
# set current train and test
test = np.array(master_fold_list[v_index])
temp = master_fold_list[:v_index] + master_fold_list[v_index + 1:]
train = np.array(helper_depack(temp))
t_test = test[:, 0]
t_train = train[:, 0]
X_test = test[:, 1:]
X_train = train[:, 1:]
# put train to ols_coefficent_prediction_lamda (X,t,lamb) and get w
w = ols_main_vector.ols_coefficent_prediction_lamda(
X_train, t_train, lamb)
# generate t_hat by calling generate_predition_vector(x,w) where w from previous step and x is test set
t_hat = ols_main_vector.generate_predition_vector(X_test, w)
# use MSE function to caluclate (t,t_hat) for error rate
socre = mean_squared_error(t_test, t_hat)
MSE_list.append(socre)
# keep track of each loop's MSE and return the mean and standard deviation across folds of the MSE.
std = statistics.stdev(MSE_list)
mean = statistics.mean(MSE_list)
print(std, mean)
return mean, std
def compare(master_trainning_x, master_trainning_t, master_testing_x,
master_testing_t, D, lamb):
result = list()
result.append(0)
for power in range(1, D + 1):
master_traiining_X = ols_main_vector.creates_predictor_matrix(
master_trainning_x, power)
w = ols_main_vector.ols_coefficent_prediction_lamda(
master_traiining_X, master_trainning_t, lamb)
t_hat = ols_main_vector.generate_predition_vector(master_testing_x, w)
error = cross.mean_squared_error(master_testing_t, t_hat)
result.append(error)
return result