refactor(qav3): factor out pairwise distance sums computation

Signed-off-by: Jérôme Benoit <jerome.benoit@piment-noir.org>

diff --git a/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py b/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
index 2468a7f575bc535a0c174311bfa892fd94fe1e49..8d2c468a6c83b494588e65b82648471fea2fb40f 100644 (file)
--- a/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
+++ b/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
@@ -704,6 +704,50 @@ class QuickAdapterRegressorV3(BaseRegressionModel):
             )
             return np.median(values)
 
+    @staticmethod
+    def _pairwise_distance_sums(
+        matrix: NDArray[np.floating],
+        metric: str,
+        *,
+        weights: Optional[NDArray[np.floating]] = None,
+        p: Optional[float] = None,
+    ) -> NDArray[np.floating]:
+        """Return, for each sample, the sum of its distances to all other samples.
+
+        Typical usage: representative (e.g., medoid) selection by taking argmin of
+        the returned vector. Function is generic and not tied to the medoid concept.
+
+        Parameters:
+        - matrix: 2D array (n_samples, n_features) assumed already normalized.
+        - metric: distance metric name accepted by scipy.spatial.distance.cdist.
+        - weights: optional weight vector (broadcast via cdist 'w' parameter) for metrics supporting it.
+        - p: optional Minkowski order when metric == 'minkowski'.
+
+        Notes:
+        - Caller must validate metric compatibility (e.g. exclude mahalanobis / seuclidean / jensenshannon here).
+        - Behavior for n_samples in {0,1} is handled to preserve previous semantics.
+        """
+        if matrix.ndim != 2:
+            raise ValueError("matrix must be 2-dimensional")
+        if matrix.shape[1] == 0:
+            raise ValueError("matrix must have at least one feature")
+        if matrix.shape[0] == 1:
+            return np.array([0.0])
+        if matrix.shape[0] < 2:
+            return np.full(matrix.shape[0], np.inf)
+        cdist_kwargs: dict[str, Any] = {}
+        if weights is not None:
+            cdist_kwargs["w"] = weights
+        if metric == "minkowski" and p is not None:
+            cdist_kwargs["p"] = p
+        pairwise_distances = sp.spatial.distance.cdist(
+            matrix,
+            matrix,
+            metric=metric,
+            **cdist_kwargs,
+        )
+        return np.sum(pairwise_distances, axis=1)
+
     def get_multi_objective_study_best_trial(
         self, namespace: str, study: optuna.study.Study
     ) -> Optional[optuna.trial.FrozenTrial]:
@@ -888,18 +932,17 @@ class QuickAdapterRegressorV3(BaseRegressionModel):
                     raise ValueError(
                         f"Unsupported label_medoid_metric: {label_medoid_metric}. Supported are euclidean/minkowski/cityblock/chebyshev/..."
                     )
-                cdist_kwargs: dict[str, Any] = {"w": np_weights}
-                if label_medoid_metric == "minkowski":
-                    cdist_kwargs["p"] = (
-                        label_p_order if label_p_order is not None else 2.0
-                    )
-                pairwise_distances = sp.spatial.distance.cdist(
-                    normalized_matrix,
+                return self._pairwise_distance_sums(
                     normalized_matrix,
-                    metric=label_medoid_metric,
-                    **cdist_kwargs,
+                    label_medoid_metric,
+                    weights=np_weights,
+                    p=(
+                        label_p_order
+                        if label_medoid_metric == "minkowski"
+                        and label_p_order is not None
+                        else None
+                    ),
                 )
-                return np.sum(pairwise_distances, axis=1)
             elif metric in {"kmeans", "kmeans2"}:
                 if n_samples == 1:
                     return np.array([0.0])
@@ -952,14 +995,17 @@ class QuickAdapterRegressorV3(BaseRegressionModel):
                 if best_cluster_indices is not None and best_cluster_indices.size > 0:
                     if label_kmeans_selection == "medoid":
                         best_cluster_matrix = normalized_matrix[best_cluster_indices]
-                        pairwise_distances = sp.spatial.distance.cdist(
-                            best_cluster_matrix,
-                            best_cluster_matrix,
-                            metric=label_kmeans_metric,
-                            **cdist_kwargs,
-                        )
-                        trial_distances[best_cluster_indices] = np.sum(
-                            pairwise_distances, axis=1
+                        trial_distances[best_cluster_indices] = (
+                            self._pairwise_distance_sums(
+                                best_cluster_matrix,
+                                label_kmeans_metric,
+                                p=(
+                                    label_p_order
+                                    if label_kmeans_metric == "minkowski"
+                                    and label_p_order is not None
+                                    else None
+                                ),
+                            )
                         )
                     elif label_kmeans_selection == "min":
                         trial_distances[best_cluster_indices] = (