def test_size_time_for_various_alphas(n, alpha_count):
d = 7
m = 3
alphas = np.ones(alpha_count)
A_tag = get_varying_d_data(d=d, n=n)
ridgecv_mes = measure_method(
RidgeCV(alphas=alphas, cv=m).fit, A_tag[:, :d], A_tag[:, d])
ridge_boost_mes = measure_method(ridgecv_boost, A_tag, alphas, m,
get_optimal_k_value(d))
lassocv_mes = measure_method(
LassoCV(alphas=alphas, cv=m).fit, A_tag[:, :d], A_tag[:, d])
lasso_boost_mes = measure_method(lassocv_boost, A_tag, alphas, m,
get_optimal_k_value(d))
elasticcv_mes = measure_method(
ElasticNetCV(alphas=alphas, cv=m, l1_ratio=RHO).fit, A_tag[:, :d],
A_tag[:, d])
elastic_boost_mes = measure_method(elasticcv_boost, A_tag, m, alphas, RHO,
get_optimal_k_value(d))
with open('SECOND_PERFORMANCE_TEST_RESULTS.txt', 'a+') as f:
f.write("ridge,{n},{alpha_count},{duration}\n".format(
n=n, alpha_count=alpha_count, duration=ridgecv_mes.duration))
f.write("ridge_boost,{n},{alpha_count},{duration}\n".format(
n=n, alpha_count=alpha_count, duration=ridge_boost_mes.duration))
f.write("lasso,{n},{alpha_count},{duration}\n".format(
n=n, alpha_count=alpha_count, duration=lassocv_mes.duration))
f.write("lasso_boost,{n},{alpha_count},{duration}\n".format(
n=n, alpha_count=alpha_count, duration=lasso_boost_mes.duration))
f.write("elastic,{n},{alpha_count},{duration}\n".format(
n=n, alpha_count=alpha_count, duration=elasticcv_mes.duration))
f.write("elastic_boost,{n},{alpha_count},{duration}\n".format(
n=n, alpha_count=alpha_count, duration=elastic_boost_mes.duration))
def test_time_for_increasing_alphas(dataset, dataset_name, alpha_count):
m = 3
alphas = np.ones(alpha_count)
A_tag = get_dataset(dataset)
n, d_tag = A_tag.shape
d = d_tag - 1
ridgecv_mes = measure_method(
RidgeCV(alphas=alphas, cv=m).fit, A_tag[:, :d], A_tag[:, d])
ridge_boost_mes = measure_method(ridgecv_boost, A_tag, alphas, m,
get_optimal_k_value(d))
lassocv_mes = measure_method(
LassoCV(alphas=alphas, cv=m).fit, A_tag[:, :d], A_tag[:, d])
lasso_boost_mes = measure_method(lassocv_boost, A_tag, alphas, m,
get_optimal_k_value(d))
elasticcv_mes = measure_method(
ElasticNetCV(alphas=alphas, cv=m, l1_ratio=RHO).fit, A_tag[:, :d],
A_tag[:, d])
elastic_boost_mes = measure_method(elasticcv_boost, A_tag, m, alphas, RHO,
get_optimal_k_value(d))
with open('THIRD_PERFORMANCE_TEST_RESULTS.txt', 'a+') as f:
f.write("ridge,{dataset_name},{alpha_count},{duration}\n".format(
dataset_name=dataset_name,
alpha_count=alpha_count,
duration=ridgecv_mes.duration))
f.write("ridge_boost,{dataset_name},{alpha_count},{duration}\n".format(
dataset_name=dataset_name,
alpha_count=alpha_count,
duration=ridge_boost_mes.duration))
f.write("lasso,{dataset_name},{alpha_count},{duration}\n".format(
dataset_name=dataset_name,
alpha_count=alpha_count,
duration=lassocv_mes.duration))
f.write("lasso_boost,{dataset_name},{alpha_count},{duration}\n".format(
dataset_name=dataset_name,
alpha_count=alpha_count,
duration=lasso_boost_mes.duration))
f.write("elastic,{dataset_name},{alpha_count},{duration}\n".format(
dataset_name=dataset_name,
alpha_count=alpha_count,
duration=elasticcv_mes.duration))
f.write(
"elastic_boost,{dataset_name},{alpha_count},{duration}\n".format(
dataset_name=dataset_name,
alpha_count=alpha_count,
duration=elastic_boost_mes.duration))
def test_linreg_datasets_accuracy(dataset):
A_tag = get_dataset(dataset)
(n, d_tag) = A_tag.shape
d = d_tag - 1
k = get_optimal_k_value(d)
x_fast = linreg_boost(A_tag, m=1, k=k)[0]
x = scipy.linalg.lstsq(A_tag[:, 0:d], A_tag[:, d])[0]
err = np.linalg.norm(x - x_fast)
print(err)
print(err / np.linalg.norm(x))
def test_lasso_datasets_accuracy(dataset):
A_tag = get_dataset(dataset)
(n, d_tag) = A_tag.shape
alphas = (0.1, 1, 10, 100, 1000)
d = d_tag - 1
m = 64
k = get_optimal_k_value(d)
result_fast = lassocv_boost(A_tag, alphas=alphas, m=m, k=k)
result = LassoCV(alphas=alphas, cv=m).fit(A_tag[:, 0:d], A_tag[:, d])
err = np.linalg.norm(result_fast.coef_ - result.coef_)
print(err)
print(err / np.linalg.norm(result.coef_))
def test_elastic_datasets_accuracy(dataset):
A_tag = get_dataset(dataset)
(n, d_tag) = A_tag.shape
alphas = (0.1, 1, 10, 100, 1000)
d = d_tag - 1
m = 2
k = get_optimal_k_value(d)
rho = 0.5
result_fast = elasticcv_boost(A_tag, alphas=alphas, m=m, k=k, rho=rho)
result = ElasticNetCV(alphas=alphas, cv=m, l1_ratio=rho).fit(A_tag[:, 0:d], A_tag[:, d])
err = np.linalg.norm(result_fast.coef_ - result.coef_)
print(err)
print(err / np.linalg.norm(result.coef_))
def test_linreg_various_values(n, d):
m = 64
A_tag = get_varying_d_data(d, n)
n, d_tag = A_tag.shape
d = d_tag - 1
linreg_mes = measure_method(scipy.linalg.lstsq, A_tag[:, :d], A_tag[:, d])
linreg_boost_mes = measure_method(linreg_boost, A_tag, m,
get_optimal_k_value(d))
with open('LINREG_PERFORMANCE_TEST_D_VALUES_very_large.txt', 'a+') as f:
f.write("linreg,{n},{d},{duration}\n".format(
n=n, d=d, duration=linreg_mes.duration))
f.write("linreg_boost,{n},{d},{duration}\n".format(
n=n, d=d, duration=linreg_boost_mes.duration))