labels_df_test = dfs_test[target]
labels_df_test.columns = [attr_name + "0", attr_name + "3"]
true_links = pd.concat([labels_df_train, labels_df_test])
true_links = true_links.loc[true_links['Relation'] == target]
true_links.drop(['Relation'], axis=1, inplace=True)
cols = true_links.columns.tolist()
inv_true_links = true_links[[cols[1], cols[0]]].copy()
inv_true_links.columns = [cols[0], cols[1]]
mask = [1] * len(rel_names_train)
pos = rel_names_train.index(target)
mask[pos:pos + 2] = [0] * 2
yhat = score(disj, 1024, target, mask)
print("done evaluation (" + str(time.time() - begtime) + "s)")
mrr = 0.0
hits10 = 0.0
hits1 = 0.0
hits3 = 0.0
ranks = []
scores = pd.merge(labels_df_test,
yhat,
left_on=[attr_name + "0", attr_name + "3"],
right_on=[attr_name + "0", attr_name + "3"],
how='left')
scores.fillna(0, inplace=True)
print("done retrieving scores (" + str(time.time() - begtime) + "s)")
names=['id', 'type'])
dfs_test = load_metadata(background_fname)
load_data(facts_fname_test, dfs_test)
labels_df_test = dfs_test[target]
labels_df_test.columns = [attr_name + "0", attr_name + "3"]
true_links = pd.concat([labels_df_train, labels_df_test])
true_links.drop(['Label'], axis=1, inplace=True)
cols = true_links.columns.tolist()
inv_true_links = true_links[[cols[1], cols[0]]].copy()
inv_true_links.columns = [cols[0], cols[1]]
yhat = score(proj, 1024)
yhat[['new_' + attr_name + "3"]] = yhat[[attr_name + "3"
]].apply(pd.to_numeric)
yhat = pd.merge(yhat, dest2type, \
left_on=['new_' + attr_name + "3"], \
right_on=['id'], \
how='left')
yhat.drop(['new_' + attr_name + "3", "id"], axis=1, inplace=True)
yhat.fillna("null", inplace=True)
print("done evaluation (" + str(time.time() - begtime) + "s)")
mrr = 0.0
hits10 = 0.0
hits1 = 0.0
hits3 = 0.0
ranks = []
dfs_test = load_metadata(background_fname)
load_data(facts_fname_test, dfs_test)
labels_df_test = dfs_test[target]
labels_df_test.columns = [attr_name + "0", attr_name + "3"]
labels_df_test['Label'] = 1.0
true_links = pd.concat([labels_df_train, labels_df_test])
true_links.drop(['Label'], axis=1, inplace=True)
cols = true_links.columns.tolist()
inv_true_links = true_links[[cols[1], cols[0]]].copy()
inv_true_links.columns = [cols[0], cols[1]]
yhat = score(disj, 1024)
print("done evaluation (" + str(time.time()-begtime) + "s)")
mrr = 0.0
hits10 = 0.0
hits1 = 0.0
hits3 = 0.0
ranks = []
for index, row in labels_df_test.iterrows():
src = row[attr_name + "0"]
dest = row[attr_name + "3"]
print("query: " + src + " " + dest)
#looking for dest; minerva evaluates this way
score4dest, rank4dest = eval(yhat, attr_name + "0", src, attr_name + "3", dest, true_links)
#score4dest_inv, rank4dest_inv = eval(yhat, attr_name + "0", src, attr_name + "3", dest, inv_true_links)
str(weights.detach().numpy()))
lnn_beta, lnn_wts, slacks = join.AND['default'].cdd()
np.set_printoptions(precision=3, suppress=True)
print("LNN beta, weights: " + \
str(np.around(lnn_beta.item(), decimals=3)) + " " + str(lnn_wts.detach().numpy()))
dfs_test = load_metadata(background_fname)
load_data(facts_fname_test, dfs_test)
labels_df_test = dfs_test[target]
labels_df_test.columns = [attr_name + "0", attr_name + "3"]
print("read test data")
#yhat = score(disj, batch_size)
yhat = score(proj, batch_size)
print("done evaluation (" + str(time.time() - begtime) + "s)")
test_countries = list(
set(labels_df_test[[attr_name + "0"
]].to_numpy().transpose()[0].tolist()))
test_regions = list(
set(labels_df_test[[attr_name + "3"
]].to_numpy().transpose()[0].tolist()))
fout = open('auc.csv', 'w')
for test_c in test_countries:
for test_r in test_regions:
check = labels_df_test.loc[
(labels_df_test[attr_name + "0"] == test_c)
& (labels_df_test[attr_name + "3"] == test_r)]