Merge pull request #525 from Hind-M/knn_dtm_overflow

Warn about overflow/nan in knn

2 files changed, 24 insertions, 0 deletions

diff --git a/src/python/gudhi/point_cloud/knn.py b/src/python/gudhi/point_cloud/knn.py
index 829bf1bf..de5844f9 100644
--- a/src/python/gudhi/point_cloud/knn.py
+++ b/src/python/gudhi/point_cloud/knn.py
@@ -8,6 +8,7 @@
 #   - YYYY/MM Author: Description of the modification
 
 import numpy
+import warnings
 
 # TODO: https://github.com/facebookresearch/faiss
 
@@ -257,6 +258,9 @@ class KNearestNeighbors:
             if ef is not None:
                 self.graph.set_ef(ef)
             neighbors, distances = self.graph.knn_query(X, k, num_threads=self.params["num_threads"])
+            with warnings.catch_warnings():
+                if not(numpy.all(numpy.isfinite(distances))):
+                    warnings.warn("Overflow/infinite value encountered while computing 'distances'", RuntimeWarning)
             # The k nearest neighbors are always sorted. I couldn't find it in the doc, but the code calls searchKnn,
             # which returns a priority_queue, and then fills the return array backwards with top/pop on the queue.
             if self.return_index:
@@ -290,6 +294,9 @@ class KNearestNeighbors:
             if self.return_index:
                 if self.return_distance:
                     distances, neighbors = mat.Kmin_argKmin(k, dim=1)
+                    with warnings.catch_warnings():
+                        if not(torch.isfinite(distances).all()):
+                            warnings.warn("Overflow/infinite value encountered while computing 'distances'", RuntimeWarning)
                     if p != numpy.inf:
                         distances = distances ** (1.0 / p)
                     return neighbors, distances
@@ -298,6 +305,9 @@ class KNearestNeighbors:
                     return neighbors
             if self.return_distance:
                 distances = mat.Kmin(k, dim=1)
+                with warnings.catch_warnings():
+                    if not(torch.isfinite(distances).all()):
+                        warnings.warn("Overflow/infinite value encountered while computing 'distances'", RuntimeWarning)
                 if p != numpy.inf:
                     distances = distances ** (1.0 / p)
                 return distances
diff --git a/src/python/test/test_dtm.py b/src/python/test/test_dtm.py
index 0a52279e..c29471cf 100755
--- a/src/python/test/test_dtm.py
+++ b/src/python/test/test_dtm.py
@@ -13,6 +13,7 @@ import numpy
 import pytest
 import torch
 import math
+import warnings
 
 
 def test_dtm_compare_euclidean():
@@ -87,3 +88,16 @@ def test_density():
     assert density == pytest.approx(expected)
     density = DTMDensity(weights=[0.5, 0.5], metric="neighbors", dim=1).fit_transform(distances)
     assert density == pytest.approx(expected)
+
+def test_dtm_overflow_warnings():
+    pts = numpy.array([[10., 100000000000000000000000000000.], [1000., 100000000000000000000000000.]])
+    impl_warn = ["keops", "hnsw"]
+
+    with warnings.catch_warnings(record=True) as w:
+        for impl in impl_warn:
+            dtm = DistanceToMeasure(2, q=10000, implementation=impl)
+            r = dtm.fit_transform(pts)
+        assert len(w) == 2
+        for i in range(len(w)):
+            assert issubclass(w[i].category, RuntimeWarning)
+            assert "Overflow" in str(w[i].message)