Merge branch 'main' into pre-comp-signature-passing

src/signature_mahalanobis_knn/mahal_distance.py

@@ -5,24 +5,41 @@
 
 
 class Mahalanobis:
-    def __init__(self):
+    def __init__(
+        self,
+        subspace_thres: float = 1e-3,
+        svd_thres: float = 1e-12,
+        zero_thres: float = 1e-15,
+    ):
         """
         After fit is called, becomes callable and intended to be used
         as a distance function in sklearn nearest neighbour.
+
+        Parameters
+        ----------
+        subspace_thres : float, optional
+            Threshold to decide whether a point is in the data subspace,
+            by default 1e-3.
+        svd_thres : float, optional
+            Threshold to decide numerical rank of the data matrix,
+            by default 1e-12.
+        zero_thres : float, optional
+            Threshold to decide whether the distance is zero,
+            by default 1e-15.
         """
-        self.Vt: np.ndarray = np.empty(
-            0
-        )  # Truncated right singular matrix transposed of the corpus
-        self.mu: np.ndarray = np.empty(0)  # Mean of the corpus
-        self.S: np.ndarray = np.empty(0)  # Truncated singular values of the corpus
-        self.subspace_thres: float = (
-            1e-3  # Threshold to decide whether a point is in the data subspace
-        )
-        self.svd_thres: float = (
-            1e-12  # Threshold to decide numerical rank of the data matrix
-        )
-        self.numerical_rank: int = -1  # Numerical rank
-        self.default_dtype = np.float64
+        self.subspace_thres: float = subspace_thres
+        self.svd_thres: float = svd_thres
+        self.zero_thres: float = zero_thres
+
+        # set the following after fit() is called - None means not fitted yet
+        # truncated right singular matrix transposed of the corpus
+        self.Vt: np.ndarray | None = None
+        # nean of the corpus
+        self.mu: np.ndarray | None = None
+        # truncated singular values of the corpus
+        self.S: np.ndarray | None = None
+        # numerical rank of the corpus
+        self.numerical_rank: int | None = None
 
     def fit(self, X: np.ndarray, y: None = None, **kwargs) -> None:  # noqa: ARG002
         """
@@ -53,6 +70,7 @@ def calc_distance(
         Vt: np.ndarray,
         S: np.ndarray,
         subspace_thres: float,
+        zero_thres: float,
     ) -> float:
         """
         Compute the variance norm between x1 and x2 using the precomputed SVD.
@@ -77,7 +95,7 @@ def calc_distance(
         """
         x = x1 - x2
         norm_x = np.linalg.norm(x)
-        if norm_x < 1e-15:
+        if norm_x < zero_thres:
             return 0.0
 
         # quantifies the amount that x is outside the row-subspace
@@ -103,8 +121,19 @@ def distance(self, x1: np.ndarray, x2: np.ndarray) -> float:
         float
             Value representing distance between x, y.
         """
+        if self.numerical_rank is None:
+            msg = "Mahalanobis distance is not fitted yet."
+            raise ValueError(msg)
+
         # ensure inputs are the right data type
         x1 = x1.astype(self.default_dtype)
         x2 = x2.astype(self.default_dtype)
 
-        return self.calc_distance(x1, x2, self.Vt, self.S, self.subspace_thres)
+        return self.calc_distance(
+            x1=x1,
+            x2=x2,
+            Vt=self.Vt,
+            S=self.S,
+            subspace_thres=self.subspace_thres,
+            zero_thres=self.zero_thres,
+        )