Update to version 0.5

.github/workflows/checks.yaml

@@ -12,9 +12,9 @@ jobs:
   pre-commit:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@v8.1.0
         with:
           enable-cache: true
       - name: Run pre-commit

.github/workflows/docs.yaml

@@ -17,14 +17,14 @@ jobs:
   build:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
       - name: Setup uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@v8.1.0
         with:
           python-version: '3.14'
           enable-cache: true
       - name: Setup just
-        uses: extractions/setup-just@v3
+        uses: extractions/setup-just@v4
       - name: Setup Khiops
         run: |
           sudo apt-get update -y && sudo apt-get install wget lsb-release -y
@@ -35,7 +35,7 @@ jobs:
           mkdir docs/_static
           just docs
       - name: Upload the package as an artifact
-        uses: actions/upload-pages-artifact@v4
+        uses: actions/upload-pages-artifact@v5
         with:
           path: ./docs/_build/html
   publish:
@@ -44,7 +44,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Setup Github Pages
-        uses: actions/configure-pages@v5
+        uses: actions/configure-pages@v6
       - name: Deploy docs to GitHub Pages
         id: deployment
-        uses: actions/deploy-pages@v4
+        uses: actions/deploy-pages@v5

.github/workflows/package.yaml

@@ -19,9 +19,9 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout sources
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
       - name: Setup uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@v8.1.0
         with:
           python-version: '3.14'
           enable-cache: true
@@ -44,7 +44,7 @@ jobs:
       id-token: write
     steps:
       - name: Download package artifact
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v8
         with:
           name: pip-package
           path: dist/
@@ -61,7 +61,7 @@ jobs:
       id-token: write
     steps:
       - name: Download package artifact
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v8
         with:
           name: pip-package
           path: dist/

.github/workflows/tests.yaml

@@ -20,16 +20,16 @@ jobs:
         os: ['ubuntu-latest', 'windows-latest', 'macos-latest']
     runs-on: ${{ matrix.os }}
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
       - name: Install uv
-        uses: astral-sh/setup-uv@v6
+        uses: astral-sh/setup-uv@v8.1.0
         with:
           python-version: ${{ matrix.python-version }}
           enable-cache: true
       - name: Setup just
-        uses: extractions/setup-just@v3
+        uses: extractions/setup-just@v4
       - name: Install Miniforge
-        uses: conda-incubator/setup-miniconda@v3
+        uses: conda-incubator/setup-miniconda@v4
         with:
           auto-update-conda: true
           miniforge-version: latest

CHANGELOG.md

@@ -1,5 +1,11 @@
 # Changelog
 
+## [0.5] - 2026-04-30
+
+### Fixed
+
+- `compute_dirac_indexes`: Incorrect result for single-bin histograms.
+
 ## [0.4] - 2026-01-05
 
 ### Fixed
@@ -23,7 +29,8 @@
 - `calibration_error` function
 - `build_reliability_diagram` function
 
-[Unreleased]: https://github.com/khiopslab/khalib/compare/0.4...HEAD
+[Unreleased]: https://github.com/khiopslab/khalib/compare/0.5...HEAD
+[0.5]: https://github.com/khiopslab/khalib/compare/0.4...0.5
 [0.4]: https://github.com/khiopslab/khalib/compare/0.3...0.4
 [0.3]: https://github.com/khiopslab/khalib/compare/0.2...0.3
 [0.2]: https://github.com/khiopslab/khalib/compare/0.1...0.2

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "khalib"
-version = "0.4"
+version = "0.5"
 description = "Classifier Calibration with Khiops"
 authors = [{ name = "Felipe Olmos", email = "luisfelipe.olmosmarchant@orange.com" }]
 requires-python = ">=3.11"

src/khalib/main.py

@@ -304,18 +304,16 @@ def from_data(
                     y_indexes = le.transform(y)
                     x_bin_indexes = (
                         np.searchsorted(breakpoints[:-1], x, side="left") - 1
-                    ).reshape(1, -1)
+                    ).reshape(-1)
                     x_bin_indexes[x_bin_indexes < 0] = 0
-                    # Note: reshape is needed for the 1-column matrix case if not the
-                    # iterator below does not work
 
                     # Compute the target frequencies
                     target_freqs = [[0 for _ in le.classes_] for _ in breakpoints[:-1]]
                     for y_score_bin_index, y_index in np.nditer(
                         [x_bin_indexes, y_indexes]
                     ):
                         target_freqs[y_score_bin_index][y_index] += 1
-                    target_freqs = [tuple(freqs) for freqs in target_freqs]
+                    target_freqs = [tuple(tf) for tf in target_freqs]
         # Otherwise there is just one interval
         else:
             # Non-informative variable: histogram with only the bin (min, max)
@@ -417,8 +415,10 @@ def __post_init__(self):
                         f"the number of classes: {len(tfreqs)} != {len(self.classes)}."
                     )
                 if sum(tfreqs) != self.freqs[i]:
-                    f"`target_freqs` at bin index {i} sums different from the bin "
-                    f"frequency: {sum(tfreqs)} != {self.freqs[i]}"
+                    raise ValueError(
+                        f"`target_freqs` at bin index {i} sums different from the bin "
+                        f"frequency: {sum(tfreqs)} != {self.freqs[i]}"
+                    )
 
         # Initialize the densities and target probabilities
         self.densities = [
@@ -797,11 +797,6 @@ def fit(self, X, y):  # noqa: N803
                 for k in range(n_classes)
             ]
 
-            calibrated_probas = np.empty(scores.shape)
-            for k, histogram in enumerate(self.histograms_):
-                calibrated_probas[:, k] = calibrate_binary(
-                    scores[:, k], histogram, only_positive=True
-                )
         return self
 
     def predict_proba(self, X):  # noqa: N803
@@ -913,7 +908,7 @@ def build_reliability_diagram(
         *Density plot only:* If the log-difference between the maximal and minimal
         positive density values is larger than 'log_plot_threshold' then the density
         plot uses a log scale in the y-axis.
-    min_density_bar_width : float, default=5.0e-03
+    min_density_bar_width : float, default=2.5e-03
         *Density plot only:* If a bin of the scores' unsupervised histogram has a width
         lower than 'min_density_bar_width' then it is plotted as having a width of
         'min_density_bar_width'.
@@ -935,7 +930,7 @@ def build_reliability_diagram(
         )
     if len(y.shape) > 1:
         raise ValueError(
-            f"'y_scores' must be a 1-D array-like object, but its shape is {y.shape}"
+            f"'y' must be a 1-D array-like object, but its shape is {y.shape}"
         )
     if dirac_threshold <= 0.0:
         raise ValueError("'dirac_threshold' must be positive")
@@ -1049,7 +1044,7 @@ def build_reliability_diagram(
             )
             if freq > 0
         ]
-        density_bar_log_range = None
+        density_bar_log_range = 0
         if density_bar_heights:
             density_bar_log_range = math.log10(max(density_bar_heights)) - math.log10(
                 min(density_bar_heights)
@@ -1078,10 +1073,19 @@ def compute_dirac_indexes(uhist, dirac_threshold):
 
     We declare a dirac mass bin if:
 
-    - it is surrounded by empty bins.
+    - it is surrounded by empty bins or if it is the only one
     - its length is less than ``dirac_threshold``
 
     """
+
+    # Early return for just one bin
+    if len(uhist.freqs) == 1:
+        if uhist.bins[0][1] - uhist.bins[0][0] < dirac_threshold:
+            return [True]
+        else:
+            return [False]
+
+    # More than one bin
     dirac_indexes = []
     if (
         len(uhist.freqs) > 1
@@ -1102,7 +1106,7 @@ def compute_dirac_indexes(uhist, dirac_threshold):
         else:
             dirac_indexes.append(False)
     if (
-        len(uhist.freqs) > 2
+        len(uhist.freqs) > 1
         and uhist.freqs[-2] == 0
         and (uhist.bins[-1][1] - uhist.bins[-1][0] < dirac_threshold)
     ):

src/tests/test_all.py

@@ -155,6 +155,19 @@ def test_find_vfind_coherence(self, y_fixtures, y_scores_fixtures):
             [histogram.find(score) for score in histogram.breakpoints],
         )
 
+    def test_find_vfind_boundary_coherence(self):
+        """Test find/vfind agree on all boundary edge cases"""
+        histogram = Histogram(
+            breakpoints=[0.0, 0.3, 0.7, 1.0],
+            freqs=[10, 20, 15],
+        )
+        # Test: below min, at each breakpoint, between bins, above max
+        edge_values = [-0.5, 0.0, 0.15, 0.3, 0.5, 0.7, 0.85, 1.0, 1.5]
+        np.testing.assert_array_equal(
+            histogram.vfind(edge_values),
+            [histogram.find(v) for v in edge_values],
+        )
+
     @pytest.mark.parametrize("y_scores_fixture", ["original", "constant"])
     @pytest.mark.parametrize("method", ["eq-freq", "eq-width", "khiops"])
     def test_manual_vs_khiops_coherence(
@@ -452,7 +465,7 @@ def test_uniform(self):
         rng = np.random.default_rng(seed=1234567)
         y_scores = rng.uniform(size=2000)
         uhist = Histogram.from_data(y_scores, use_finest=True)
-        assert compute_dirac_indexes(uhist, 1e-06) == [False, False]
+        assert compute_dirac_indexes(uhist, 1e-06) == [False]
 
     def test_dirac(self):
         y_scores = np.array([0.2] * 250 + [0.5] * 250 + [0.9] * 500)