Support bit_hamming and bit_jaccard distances.

umap/umap_.py

@@ -63,6 +63,7 @@
     "cosine": 2,
     "hellinger": 1,
     "jaccard": 1,
+    "bit_jaccard": 1,
     "dice": 1,
 }
 
@@ -2351,8 +2352,10 @@ def fit(self, X, y=None, force_all_finite=True):
                                    - 'allow-nan': accepts only np.nan and pd.NA values in array.
                                      Values cannot be infinite.
         """
-
-        X = check_array(X, dtype=np.float32, accept_sparse="csr", order="C", force_all_finite=force_all_finite)
+        if self.metric in ("bit_hamming", "bit_jaccard"):
+            X = check_array(X, dtype=np.uint8, order="C", force_all_finite=force_all_finite)
+        else:
+            X = check_array(X, dtype=np.float32, accept_sparse="csr", order="C", force_all_finite=force_all_finite)
         self._raw_data = X
 
         # Handle all the optional arguments, setting default
@@ -2926,7 +2929,10 @@ def transform(self, X, force_all_finite=True):
                 "Transform unavailable when model was fit with only a single data sample."
             )
         # If we just have the original input then short circuit things
-        X = check_array(X, dtype=np.float32, accept_sparse="csr", order="C", force_all_finite=force_all_finite)
+        if self.metric in ("bit_hamming", "bit_jaccard"):
+            X = check_array(X, dtype=np.uint8, order="C", force_all_finite=force_all_finite)
+        else:
+            X = check_array(X, dtype=np.float32, accept_sparse="csr", order="C", force_all_finite=force_all_finite)
         x_hash = joblib.hash(X)
         if x_hash == self._input_hash:
             if self.transform_mode == "embedding":
@@ -3297,7 +3303,10 @@ def _output_dist_only(x, y, *kwds):
         return inv_transformed_points
 
     def update(self, X, force_all_finite=True):
-        X = check_array(X, dtype=np.float32, accept_sparse="csr", order="C", force_all_finite=force_all_finite)
+        if self.metric in ("bit_hamming", "bit_jaccard"):
+            X = check_array(X, dtype=np.uint8, order="C", force_all_finite=force_all_finite)
+        else:
+            X = check_array(X, dtype=np.float32, accept_sparse="csr", order="C", force_all_finite=force_all_finite)
         random_state = check_random_state(self.transform_seed)
         rng_state = random_state.randint(INT32_MIN, INT32_MAX, 3).astype(np.int64)