BUG - Fixed a bug where no_label_sample_selection would always assign the default value due to an incorrect condition in BNSA.

aisp/__init__.py

@@ -26,10 +26,12 @@
 https://ais-package.github.io/docs/intro
 """
 
+from importlib.metadata import version
+
 from . import csa
 from . import ina
 from . import nsa
 
 __author__ = "AISP Development Team"
-__version__ = "0.5.1"
+__version__ = version('aisp')
 __all__ = ["csa", "nsa", "ina"]

aisp/base/core/_base.py

@@ -3,6 +3,7 @@
 
 from inspect import signature
 
+
 class Base:
     """
     Generic base class for models with a common interface.
@@ -46,4 +47,4 @@ def get_params(self, deep: bool = True) -> dict:  # pylint: disable=W0613
             key: getattr(self, key)
             for key, _ in signature(self.__init__).parameters.items()
             if key != "self" and not key.startswith("_") and hasattr(self, key)
-            }
+        }

aisp/csa/_ai_recognition_sys.py

@@ -61,12 +61,10 @@ class AIRS(BaseClassifier):
         Distance metric used to compute affinity between cells and samples.
     seed : int
         Seed for the random generation of detector values. Defaults to None.
-
-    **kwargs
-        p : float
-            This parameter stores the value of ``p`` used in the Minkowski distance. The default
-            is ``2``, which represents normalized Euclidean distance.\
-            Different values of p lead to different variants of the Minkowski Distance.
+    p : float
+        This parameter stores the value of ``p`` used in the Minkowski distance. The default
+        is ``2``, which represents normalized Euclidean distance.\
+        Different values of p lead to different variants of the Minkowski Distance.
 
     Attributes
     ----------
@@ -153,7 +151,7 @@ def __init__(
 
         self.metric = sanitize_choice(metric, ["manhattan", "minkowski"], "euclidean")
 
-        self.p = p
+        self.p: float = p
 
         self._cells_memory: Optional[Dict[str | int, list[BCell]]] = None
         self._all_class_cell_vectors: Optional[List[Tuple[Any, np.ndarray]]] = None

aisp/ina/_ai_network.py

@@ -71,11 +71,10 @@ class AiNet(BaseClusterer):
     use_mst_clustering : bool, default=True
         If ``True``, performs clustering with **Minimum Spanning Tree** (MST). If ``False``,
         does not perform clustering and predict returns None.
-    **kwargs
-        p : float
-            This parameter stores the value of ``p`` used in the Minkowski distance. The default
-            is ``2``, which represents normalized Euclidean distance.\
-            Different values of p lead to different variants of the Minkowski Distance.
+    p : float
+        This parameter stores the value of ``p`` used in the Minkowski distance. The default
+        is ``2``, which represents normalized Euclidean distance.\
+        Different values of p lead to different variants of the Minkowski Distance.
 
     Attributes
     ----------
@@ -139,7 +138,7 @@ def __init__(
         metric: MetricType = "euclidean",
         seed: Optional[int] = None,
         use_mst_clustering: bool = True,
-        **kwargs,
+        p: float = 2.0
     ):
         self.N: int = sanitize_param(N, 50, lambda x: x > 0)
         self.n_clone: int = sanitize_param(n_clone, 10, lambda x: x > 0)
@@ -175,7 +174,7 @@ def __init__(
             metric, ["euclidean", "manhattan", "minkowski"], "euclidean"
         )
 
-        self.p: np.float64 = np.float64(kwargs.get("p", 2.0))
+        self.p: float = p
         self._metric_params = {}
         if self.metric == "minkowski":
             self._metric_params["p"] = self.p

aisp/nsa/_binary_negative_selection.py

@@ -118,7 +118,7 @@ def __init__(
         self.no_label_sample_selection: str = sanitize_param(
             no_label_sample_selection,
             "max_average_difference",
-            lambda x: x == "nearest_difference",
+            lambda x: x == "max_nearest_difference",
         )
 
         self.classes: Optional[npt.NDArray] = None
@@ -308,7 +308,7 @@ def _assign_class_to_non_self_sample(self, line: npt.NDArray, c: list):
             distances = np.mean(line != self._detectors[_class_], axis=1)
             # Assign the label to the class with the greatest distance from
             # the nearest detector.
-            if self.no_label_sample_selection == "nearest_difference":
+            if self.no_label_sample_selection == "max_nearest_difference":
                 class_differences[_class_] = distances.min()
             # Or based on the greatest distance from the average distances of the detectors.
             else:

aisp/nsa/_negative_selection.py

@@ -2,7 +2,7 @@
 
 from __future__ import annotations
 
-from typing import Any, Dict, Literal, Optional, Union, List
+from typing import Dict, Literal, Optional, Union, List
 
 import numpy as np
 import numpy.typing as npt
@@ -57,20 +57,17 @@ class RNSA(BaseClassifier):
         * ``'default-NSA'``: Default algorithm with fixed radius.
         * ``'V-detector'``: This algorithm is based on the article Ji & Dasgupta (2004) [1]_
             and uses a variable radius for anomaly detection in feature spaces.
-
-    **kwargs : dict
-        Additional parameters. The following arguments are recognized:  
-        * non_self_label : str, default='non-self'
-            This variable stores the label that will be assigned when the data has only one
-            output class, and the sample is classified as not belonging to that class.
-        * cell_bounds : bool, default=False
-            If set to ``True``, this option limits the generation of detectors to the space
-            within the plane between 0 and 1. This means that any detector whose radius exceeds
-            this limit is discarded, this variable is only used in the ``V-detector`` algorithm.
-        * p : float, default=2
-            This parameter stores the value of ``p`` used in the Minkowski distance. The default
-            is ``2``, which represents Euclidean distance. Different values of p lead
-            to different variants of the Minkowski Distance.
+    * non_self_label : str, default='non-self'
+        This variable stores the label that will be assigned when the data has only one
+        output class, and the sample is classified as not belonging to that class.
+    * cell_bounds : bool, default=False
+        If set to ``True``, this option limits the generation of detectors to the space
+        within the plane between 0 and 1. This means that any detector whose radius exceeds
+        this limit is discarded, this variable is only used in the ``V-detector`` algorithm.
+    * p : float, default=2
+        This parameter stores the value of ``p`` used in the Minkowski distance. The default
+        is ``2``, which represents Euclidean distance. Different values of p lead
+        to different variants of the Minkowski Distance.
 
     Attributes
     ----------
@@ -142,7 +139,9 @@ def __init__(
         max_discards: int = 1000,
         seed: Optional[int] = None,
         algorithm: Literal["default-NSA", "V-detector"] = "default-NSA",
-        **kwargs: Any,
+        p: float = 2.0,
+        non_self_label: str = 'non-self',
+        cell_bounds: bool = False
     ):
         self.metric: str = sanitize_choice(
             metric, ["manhattan", "minkowski"], "euclidean"
@@ -160,9 +159,9 @@ def __init__(
         )
         self.max_discards: int = sanitize_param(max_discards, 1000, lambda x: x > 0)
 
-        self.p: np.float64 = np.float64(kwargs.get("p", 2))
-        self.cell_bounds: bool = bool(kwargs.get("cell_bounds", False))
-        self.non_self_label: str = str(kwargs.get("non_self_label", "non-self"))
+        self.p: float = p
+        self.cell_bounds: bool = cell_bounds
+        self.non_self_label: str = non_self_label
 
         self._detectors: Optional[Dict[str | int, list[Detector]]] = None
         self.classes: Optional[npt.NDArray] = None

aisp/nsa/tests/test_negative_selection.py

@@ -103,7 +103,7 @@ def test_fit_raises_max_discards_error(self, rnsa_data):
     def test_predict_raises_feature_dimension_mismatch(self, rnsa_data):
         """Should raise FeatureDimensionMismatch when prediction input has wrong dimensions."""
         X, y, seed = rnsa_data
-        model = RNSA(N=1000, aff_thresh=0.2, seed=seed)
+        model = RNSA(N=1000, seed=seed)
         model.fit(X, y, verbose=False)
         x_invalid = np.random.rand(5, 5)
         with pytest.raises(FeatureDimensionMismatch):

aisp/utils/display.py

@@ -161,8 +161,6 @@ class ProgressTable(TableFormatter):
 
     def __init__(self, headers: Mapping[str, int], verbose: bool = True) -> None:
         super().__init__(headers)
-        if not headers or not isinstance(headers, Mapping):
-            raise ValueError("'headers' must be a non-empty dictionary.")
         self.verbose: bool = verbose
         self.headers: Mapping[str, int] = headers
         self._ascii_only = not _supports_box_drawing()

examples/en/classification/AIRS/example_with_randomly_generated_dataset-en.ipynb

@@ -204,7 +204,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": ".venv",
+   "display_name": ".venv (3.12.2)",
    "language": "python",
    "name": "python3"
   },
@@ -218,7 +218,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.11.4"
+   "version": "3.12.2"
   },
   "orig_nbformat": 4
  },