Skip to content

fix: handle zero terms in Canberra distance #5574

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
LOGO127 wants to merge 1 commit into
base: main
Choose a base branch
from
+30 −1
Open

fix: handle zero terms in Canberra distance #5574

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion faiss/utils/simd_impl/distances_autovec-inl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -208,7 +208,10 @@ float VectorDistance<METRIC_Canberra, SL>::operator()(
float accu = 0;
for (size_t i = 0; i < this->d; i++) {
float xi = x[i], yi = y[i];
accu += fabs(xi - yi) / (fabs(xi) + fabs(yi));
float denominator = fabs(xi) + fabs(yi);
if (denominator != 0.0f) {
accu += fabs(xi - yi) / denominator;
}
}
return accu;
}
Expand Down
26 changes: 26 additions & 0 deletions tests/test_extra_distances.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,32 @@ def test_canberra(self):
scipy.spatial.distance.canberra, faiss.METRIC_Canberra
)

def test_canberra_zero_denominator(self):
xq = np.array([[0.0, 0.0]], dtype="float32")
yb = np.array(
[
[0.0, 0.0],
[1.0, 0.0],
[0.0, 2.0],
],
dtype="float32",
)
expected = np.array([[0.0, 1.0, 1.0]], dtype="float32")

distances = faiss.pairwise_distances(
xq, yb, faiss.METRIC_Canberra
)
self.assertTrue(np.all(np.isfinite(distances)))
self.assertTrue(np.allclose(distances, expected))

index = faiss.IndexFlat(2, faiss.METRIC_Canberra)
index.add(yb)
distances, labels = index.search(xq, 3)
self.assertTrue(np.all(np.isfinite(distances)))
self.assertNotIn(-1, labels[0])
self.assertEqual(set(labels[0].tolist()), {0, 1, 2})
self.assertTrue(np.allclose(np.sort(distances[0]), expected[0]))

def test_braycurtis(self):
self.run_simple_dis_test(
scipy.spatial.distance.braycurtis, faiss.METRIC_BrayCurtis
Expand Down