deskit 1.2.4__tar.gz → 1.2.6__tar.gz

{deskit-1.2.4/src/deskit.egg-info → deskit-1.2.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: deskit
-Version: 1.2.4
+Version: 1.2.6
 Summary: A Python library for Dynamic Ensemble Selection
 Author: Tikhon Vodyanov
 License-Expression: MIT
@@ -146,6 +146,17 @@ weights = router.predict_weights(X_test, temperature=0.1)
 
 ---
 
+### Hyperparameter tuning
+
+`deskit` supports hyperparameter tuning of KNN-based methods using Leave One Out (LOO) tuning.
+This technique excludes the closest data point with a negligent distance from the one being used in predict(), allowing for `deskit` methods to be tuned from the same DSEL set it was fit on and removing the need for cross-validation as it sees the entire set throughout the tuning process.
+
+```python
+results = router.predict(X_val, val_preds, loo=True)
+```
+
+---
+
 ## Why deskit?
 
 Most DES libraries are tied to scikit-learn. deskit only ever sees a numpy

{deskit-1.2.4 → deskit-1.2.6}/README.md

@@ -115,6 +115,17 @@ weights = router.predict_weights(X_test, temperature=0.1)
 
 ---
 
+### Hyperparameter tuning
+
+`deskit` supports hyperparameter tuning of KNN-based methods using Leave One Out (LOO) tuning.
+This technique excludes the closest data point with a negligent distance from the one being used in predict(), allowing for `deskit` methods to be tuned from the same DSEL set it was fit on and removing the need for cross-validation as it sees the entire set throughout the tuning process.
+
+```python
+results = router.predict(X_val, val_preds, loo=True)
+```
+
+---
+
 ## Why deskit?
 
 Most DES libraries are tied to scikit-learn. deskit only ever sees a numpy

{deskit-1.2.4 → deskit-1.2.6}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "deskit"
-version = "1.2.4"
+version = "1.2.6"
 description = "A Python library for Dynamic Ensemble Selection"
 readme = "README.md"
 license = "MIT"

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/dewsi.py

@@ -38,13 +38,9 @@ class DEWSI(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
-
-    Notes
-    -----
-    predict() and predict_weights() accept an optional ``loo=True`` keyword
-    for hyperparameter tuning directly on the DSEL this model was fit on:
-    it excludes each query point's own occurrence from its neighborhood so
-    it doesn't trivially neighbor itself at distance 0.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/dewsiv.py

@@ -40,6 +40,9 @@ class DEWSIV(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,
@@ -94,7 +97,7 @@ class DEWSIV(KNNBase):
              (0.5 if self.mode == 'min' else 1.0))
         th = threshold if threshold is not None else self.threshold
 
-        distances, indices = self.model._kneighbors(x, k=k, loo=loo)                # both (batch, k)
+        distances, indices = self._kneighbors(x, k=k, loo=loo)                # both (batch, k)
 
         # Inverse-distance weights
         inv_dist   = 1.0 / np.maximum(distances, 1e-8)               # (batch, k)

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/dewst.py

@@ -43,6 +43,9 @@ class DEWST(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,
@@ -97,7 +100,7 @@ class DEWST(KNNBase):
         th    = threshold    if threshold    is not None else self.threshold
         r2_th = r2_threshold if r2_threshold is not None else self.r2_threshold
 
-        distances, indices = self.model._kneighbors(x, k=k, loo=loo)          # (batch, k)
+        distances, indices = self._kneighbors(x, k=k, loo=loo)          # (batch, k)
         k = distances.shape[1]
 
         # Inverse-distance weights

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/dewsu.py

@@ -34,6 +34,9 @@ class DEWSU(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,
@@ -75,7 +78,7 @@ class DEWSU(KNNBase):
              (0.5 if self.mode == 'min' else 1.0))
         th = threshold if threshold is not None else self.threshold
 
-        _, indices = self.model._kneighbors(x, k=k, loo=loo)                        # (batch, k)
+        _, indices = self._kneighbors(x, k=k, loo=loo)                        # (batch, k)
 
         # Average each model's scores over the K neighbors
         avg_scores = self.matrix[indices].mean(axis=1)               # (batch, n_models)

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/dewsv.py

@@ -40,6 +40,9 @@ class DEWSV(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability. 
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,
@@ -94,7 +97,7 @@ class DEWSV(KNNBase):
              (0.5 if self.mode == 'min' else 1.0))
         th = threshold if threshold is not None else self.threshold
 
-        _, indices = self.model._kneighbors(x, k=k, loo=loo)                        # (batch, k)
+        _, indices = self._kneighbors(x, k=k, loo=loo)                        # (batch, k)
 
         # Uniform average of each model's scores over K neighbors
         neighbor_scores = self.matrix[indices]                        # (batch, k, n_models)

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/knorae.py

@@ -29,6 +29,9 @@ class KNORAE(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,
@@ -64,7 +67,7 @@ class KNORAE(KNNBase):
         th       = threshold if threshold is not None else self.threshold
         n_models = len(self.models)
 
-        _, indices      = self.model._kneighbors(x, k=k, loo=loo)
+        _, indices      = self._kneighbors(x, k=k, loo=loo)
         k               = indices.shape[1]
         neighbor_scores = self.matrix[indices]                        # (batch, k, n_models)
 

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/knoraiu.py

@@ -30,6 +30,9 @@ class KNORAIU(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,
@@ -64,7 +67,7 @@ class KNORAIU(KNNBase):
         """
         th = threshold if threshold is not None else self.threshold
 
-        distances, indices = self.model._kneighbors(x, k=k, loo=loo)                # both (batch, k)
+        distances, indices = self._kneighbors(x, k=k, loo=loo)                # both (batch, k)
         neighbor_scores    = self.matrix[indices]                     # (batch, k, n_models)
 
         # Normalize per neighbor: best model = 1.0, worst = 0.0

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/knorau.py

@@ -30,6 +30,9 @@ class KNORAU(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,
@@ -64,7 +67,7 @@ class KNORAU(KNNBase):
         """
         th = threshold if threshold is not None else self.threshold
 
-        _, indices      = self.model._kneighbors(x, k=k, loo=loo)
+        _, indices      = self._kneighbors(x, k=k, loo=loo)
         neighbor_scores = self.matrix[indices]                        # (batch, k, n_models)
 
         # Normalize per neighbor: best model = 1.0, worst = 0.0

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/lwsei.py

@@ -22,6 +22,9 @@ class LWSEI(PredictBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, k=10, preset='balanced', distance_metric='euclidean', **kwargs):
@@ -87,7 +90,7 @@ class LWSEI(PredictBase):
         n_models   = len(self.models)
         uniform    = np.full(n_models, 1.0 / n_models)
 
-        distances, indices = self._finder._kneighbors(x, k=k, loo=loo)              # (batch, k)
+        distances, indices = self._kneighbors(x, k=k, loo=loo)              # (batch, k)
         weights_out        = np.empty((batch_size, n_models))
 
         for b in range(batch_size):

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/lwseu.py

@@ -22,6 +22,9 @@ class LWSEU(PredictBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, k=10, preset='balanced', distance_metric='euclidean', **kwargs):
@@ -87,7 +90,7 @@ class LWSEU(PredictBase):
         n_models   = len(self.models)
         uniform    = np.full(n_models, 1.0 / n_models)
 
-        distances, indices = self._finder._kneighbors(x, k=k, loo=loo)              # (batch, k)
+        distances, indices = self._kneighbors(x, k=k, loo=loo)              # (batch, k)
         weights_out        = np.empty((batch_size, n_models))
 
         for b in range(batch_size):

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/des/ola.py

@@ -25,6 +25,9 @@ class OLA(KNNBase):
     distance_metric : str
         Distance function to use for neighbor search. Default: 'euclidean'. See
         neighbors.list_distance_metrics() for all options and per-backend availability.
+    loo: bool
+        Enables Leave One Out (LOO) for hyperparameter tuning on the DSEL set. Default: 'false'.
+        Ignores closest neighbor with a negligible distance to avoid overfitting.
     """
 
     def __init__(self, task, metric='mae', mode='min', k=10,
@@ -63,7 +66,7 @@ class OLA(KNNBase):
         """
         batch_size = x.shape[0]
 
-        _, indices  = self.model._kneighbors(x, k=k, loo=loo)
+        _, indices  = self._kneighbors(x, k=k, loo=loo)
         avg_scores  = self.matrix[indices].mean(axis=1)               # (batch, n_models)
         best_indices = np.argmax(avg_scores, axis=1)
 

{deskit-1.2.4 → deskit-1.2.6}/src/deskit/neighbors.py

@@ -4,38 +4,6 @@ import warnings
 # FAISS IVF k-means needs at least this many training samples per cell to converge.
 _FAISS_MIN_SAMPLES_PER_CELL = 40
 
-
-# ---------------------------------------------------------------------------
-# Distance metric registry
-# ---------------------------------------------------------------------------
-
-# Metrics supported by each backend.
-# 'euclidean' is the universal default and always available.
-#
-# Choosing a distance metric:
-#   euclidean       – The standard L2 norm. Best default for most tabular data.
-#   manhattan       – L1 norm (sum of absolute differences). More robust to
-#                    outliers and tends to work better in moderately high-
-#                    dimensional spaces because it doesn't square large diffs.
-#   chebyshev       – L∞ norm (maximum absolute difference across features).
-#                    Useful when a single feature dominating the distance is
-#                    acceptable; common in game-grid / chess-style problems.
-#   minkowski       – Generalisation of L1/L2 (controlled by p). p=1 →
-#                    manhattan, p=2 → euclidean. Use when you want to tune
-#                    between them.
-#   cosine          – Angle between vectors, ignoring magnitude. Excellent for
-#                    embeddings (text, image, audio) where direction matters
-#                    more than raw scale.
-#   canberra        – Weighted L1. Sensitive to small values near zero.
-#   braycurtis      – Normalised L1 bounded to [0,1]. Common in ecology.
-#   jensenshannon   – Symmetric KL divergence on probability distributions.
-#                    Requires non-negative vectors. Supported by FAISS flat/
-#                    HNSW/GPU indices natively.
-#   dot             – Raw inner/dot product. Not a true metric; distances are
-#                    not comparable across queries. Use for max inner-product
-#                    search (recommendation systems). Prefer 'cosine' for
-#                    normalised embeddings.
-
 # Metrics that every backend supports natively.
 _UNIVERSAL_METRICS = {'euclidean', 'manhattan', 'chebyshev', 'minkowski', 'cosine'}
 

{deskit-1.2.4 → deskit-1.2.6/src/deskit.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: deskit
-Version: 1.2.4
+Version: 1.2.6
 Summary: A Python library for Dynamic Ensemble Selection
 Author: Tikhon Vodyanov
 License-Expression: MIT
@@ -146,6 +146,17 @@ weights = router.predict_weights(X_test, temperature=0.1)
 
 ---
 
+### Hyperparameter tuning
+
+`deskit` supports hyperparameter tuning of KNN-based methods using Leave One Out (LOO) tuning.
+This technique excludes the closest data point with a negligent distance from the one being used in predict(), allowing for `deskit` methods to be tuned from the same DSEL set it was fit on and removing the need for cross-validation as it sees the entire set throughout the tuning process.
+
+```python
+results = router.predict(X_val, val_preds, loo=True)
+```
+
+---
+
 ## Why deskit?
 
 Most DES libraries are tied to scikit-learn. deskit only ever sees a numpy