def test_minimax_normalization_ones():
_tca = tca_plus.TCAPlus(test_set_X=pd.DataFrame(np.ones((5, 5))),
test_set_domain=1,
train_pool_X=pd.DataFrame(np.ones((5, 5))),
train_pool_y=None,
train_pool_domain=[0, 0, 0, 1, 1],
Base_Classifier=None)
print(_tca.minmax_normalization())
def test_distance():
_tca = tca_plus.TCAPlus(test_set_X=None,
test_set_domain=None,
train_pool_X=None,
train_pool_y=None,
train_pool_domain=None,
Base_Classifier=None)
dist_ = _tca.compute_distance_set(np.random.rand(4, 50))
print(_tca.compute_dcv(dist_, 4))
def test_dcv_sim_vector():
_tca = tca_plus.TCAPlus(test_set_X=None,
test_set_domain=None,
train_pool_X=None,
train_pool_y=None,
train_pool_domain=None,
Base_Classifier=None)
dist1 = _tca.compute_distance_set(np.random.rand(4, 50))
dist2 = _tca.compute_distance_set(np.random.rand(4, 50))
dcv_1, dcv_2 = _tca.compute_dcv(dist1, 4), _tca.compute_dcv(dist2, 4)
print(_tca.compute_dist_similarity(dcv_1, dcv_2))
def test_full():
X_test = pd.DataFrame(np.random.rand(5, 5))
X_train = test_set_X = pd.DataFrame(np.ones((5, 5)))
_tca = tca_plus.TCAPlus(X_test,
test_set_domain=1,
train_pool_X=X_train,
train_pool_y=None,
train_pool_domain=[0, 0, 0, 1, 1],
Base_Classifier=None)
dist1 = _tca.compute_distance_set(X_test)
dist2 = _tca.compute_distance_set(X_train)
dcv_1, dcv_2 = _tca.compute_dcv(dist1, 4), _tca.compute_dcv(dist2, 4)
print(_tca.minmax_normalization())
def test_similarity():
_tca = tca_plus.TCAPlus(test_set_X=None,
test_set_domain=None,
train_pool_X=None,
train_pool_y=None,
train_pool_domain=None,
Base_Classifier=None)
assert _tca.compute_comp_similarity(905,
238) == tca_plus.SimLevel.MUCH_LESS
assert _tca.compute_comp_similarity(60, 60) == tca_plus.SimLevel.SAME
assert _tca.compute_comp_similarity(59, 60) == tca_plus.SimLevel.SAME
assert _tca.compute_comp_similarity(100,
125) == tca_plus.SimLevel.SLIGHTLY_MORE
assert _tca.compute_comp_similarity(100, 39) == tca_plus.SimLevel.MUCH_LESS
def test_normalization():
source_X = pd.DataFrame(np.random.rand(5, 5))
target_X = pd.DataFrame(np.random.rand(5, 5))
_tca = tca_plus.TCAPlus(test_set_X=None,
test_set_domain=1,
train_pool_X=None,
train_pool_y=None,
train_pool_domain=[0, 0, 0, 1, 1],
Base_Classifier=None)
#print(_tca.minmax_normalization(source_X, target_X))
print(_tca.zscore_normalization(source_X, target_X))
print(_tca.zscore_source_normalization(source_X, target_X))
print(_tca.zscore_target_normalization(source_X, target_X))
join="inner").reset_index(drop=True)
#============================================================
train_pool_domain = [
1 if e < len(X_train) else 0
for e in range(len(X_train_join))
]
except:
print("Error while preparing data")
continue
#============================================================
#print("TCA+")
try:
_tca = tca_plus.TCAPlus(
test_set_domain=0,
train_pool_domain=train_pool_domain,
test_set_X=X_test,
train_pool_X=X_train_join,
train_pool_y=y_train_join,
Base_Classifier=LogisticRegression)
confidence, y_p = _tca.train_filter_test()
accuracy, precision, recall, f1_score = calculate_results(
y_test, y_p)
#Store results
data = [
'TCA+', accuracy, precision, recall, f1_score, source,
target
]
performance_data.append(data)
except:
print("Error while running TCA+")
test_set_y = test_set.loc[:, ["label"]].reset_index(drop=True)
# gather all non-test indexes
train_pool = all_instances.iloc[
all_instances.index.difference(test_set.index), ]
train_pool_X = train_pool.loc[:, ["x_coord", "y_coord"]].reset_index(drop=True)
train_pool_y = train_pool["label"].reset_index(drop=True)
train_pool_domain = train_pool.domain_index
# We don't have much training data, but we got some predictions with confidence levels!
transfer_learners = [
tca_plus.TCAPlus(test_set_X=test_set_X,
test_set_domain=test_set_domain,
train_pool_X=train_pool_X,
train_pool_y=train_pool_y,
train_pool_domain=train_pool_domain,
Base_Classifier=RandomForestClassifier,
rand_seed=RAND_SEED),
tca.TCA(test_set_X=test_set_X,
test_set_domain=test_set_domain,
train_pool_X=train_pool_X,
train_pool_y=train_pool_y,
train_pool_domain=train_pool_domain,
Base_Classifier=RandomForestClassifier,
rand_seed=RAND_SEED),
tl_baseline.Source_Baseline(test_set_X=test_set_X,
test_set_domain=test_set_domain,
train_pool_X=train_pool_X,
train_pool_y=train_pool_y,
train_pool_domain=train_pool_domain,