pyFOCI 0.2.2__tar.gz → 0.3.0__tar.gz

{pyfoci-0.2.2 → pyfoci-0.3.0}/.github/workflows/deploy-gh-pages.yml

@@ -16,7 +16,7 @@ jobs:
       pages: write
 
     steps:
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@v7
       with:
         fetch-depth: 0
     - uses: prefix-dev/setup-pixi@v0.9.6

{pyfoci-0.2.2 → pyfoci-0.3.0}/.github/workflows/lint.yml

@@ -12,7 +12,7 @@ jobs:
   build:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - uses: prefix-dev/setup-pixi@v0.9.6
         with:
           pixi-version: v0.68.1

{pyfoci-0.2.2 → pyfoci-0.3.0}/.github/workflows/python-app.yml

@@ -15,7 +15,7 @@ jobs:
         os: [windows-latest, ubuntu-latest, macos-latest]
     runs-on: ${{ matrix.os }}
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - uses: prefix-dev/setup-pixi@v0.9.6
         with:
           pixi-version: v0.68.1

{pyfoci-0.2.2 → pyfoci-0.3.0}/.github/workflows/release.yml

@@ -12,7 +12,7 @@ jobs:
 
     steps:
       - name: Checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@v7
 
       - name: Install Pixi
         uses: prefix-dev/setup-pixi@v0.9.6
@@ -85,7 +85,7 @@ jobs:
     if: startsWith(github.ref, 'refs/tags/')
     steps:
       - name: Checkout
-        uses: actions/checkout@v6   # needed so we can read CHANGELOG.md
+        uses: actions/checkout@v7   # needed so we can read CHANGELOG.md
 
       - name: Download artifacts
         uses: actions/download-artifact@v8
@@ -122,14 +122,14 @@ jobs:
 
       - name: Create release (from CHANGELOG)
         if: steps.changelog.outputs.body_path != ''
-        uses: softprops/action-gh-release@v2
+        uses: softprops/action-gh-release@v3
         with:
           body_path: release_body.md
           files: dist/*
 
       - name: Create release (auto-generated notes fallback)
         if: steps.changelog.outputs.use_auto_notes == 'true'
-        uses: softprops/action-gh-release@v2
+        uses: softprops/action-gh-release@v3
         with:
           generate_release_notes: true
           files: dist/*

{pyfoci-0.2.2 → pyfoci-0.3.0}/CHANGELOG.md

@@ -5,16 +5,31 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/2.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.3.0] - 2026-06-26
+
+### Added
+- Introduced an alternative nearest neighbors selection algorithm, similar to the R reference implementation.
+  It is exposed with the new keyword-only `FOCISelector` parameter `nn_strategy="grouping"`.
+  The original algorithm remains available with `nn_strategy="radius"`.
+
+## Changed
+- Made `nn_strategy="grouping"` the new default, because it is faster.
+
+## [0.2.3] - 2026-06-23
+
+### Changed
+- Updated Action for creating GitHub releases.
+
 ## [0.2.2] - 2026-06-17
 
 ### Added
-- There is now a changelog, which is also used for the Github releases.
+- There is now a changelog, which is also used for the GitHub releases.
 
 ## [0.2.1] - 2026-06-17
 
 ### Changed
 - Input features are now N(0,1)-normalized by default.
-  This can be switched off with the new parameter `standardize` by setting it to `Ǹone`.
+  This can be switched off with the new parameter `standardize` by setting it to `None`.
 
 ## [0.2.0] - 2026-06-16
 

{pyfoci-0.2.2/pyFOCI.egg-info → pyfoci-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyFOCI
-Version: 0.2.2
+Version: 0.3.0
 Summary: A template for scikit-learn compatible packages.
 Author-email: Robert Pollak <robert.pollak@jku.at>
 License-Expression: BSD-3-Clause

{pyfoci-0.2.2 → pyfoci-0.3.0}/dev-readme.txt

@@ -9,6 +9,7 @@ pixi run -e test test
 # Also examine the resulting coverage.xml.
 
 pixi run -e doc build-doc
+# Also examine the resulting example plot.
 
 Commit to a dev branch.
 
@@ -17,7 +18,7 @@ pixi run build
 
 2. CI:
 
-Push dev branch to Github.
+Push dev branch to GitHub.
 
 Create pull request to trigger CI tests.
 
@@ -26,13 +27,20 @@ Create pull request to trigger CI tests.
 printf 'Fast-forwarding main to dev: '; git merge-base --is-ancestor main dev && git branch -f main dev && echo OK || echo ERROR
 
 git push origin main:main
-# By this, The Github repo setting "Automatically delete head branches"
+# By this, The GitHub repo setting "Automatically delete head branches"
 # deletes the dev branch in the pull request, for easier creation of the next pull request from dev.
 
+Update the [Unreleased] changelog section
+
 
 Releasing:
 ==========
 
-Following [Semantic Versioning](https://semver.org/), add a tag like "v0.1.2" to a commit and push it.
-A Github action will then publish this Release to PyPI.
+Select a version like `0.1.2`, following [Semantic Versioning](https://semver.org/).
+
+Update the changelog, commit with "Release 0.1.2".
+
+Tag the commit as "v0.1.2" and push it, including the tag.
+
+GitHub actions will then update the online docs and publish this release on GitHub and PyPI.
 

{pyfoci-0.2.2 → pyfoci-0.3.0}/pyFOCI/_foci.py

@@ -51,7 +51,78 @@ def _rank_max(y):
     return ranks
 
 
-def _Tn(X_sub, y_rank, random_state):
+def _nn_radius_based(X_sub, random_state):
+    """Radius-based NN selection with random tie breaking."""
+    X_sub = np.asarray(X_sub)
+    n = X_sub.shape[0]
+
+    # Fit NN on X_sub
+    nbrs = sklearn.neighbors.NearestNeighbors(n_neighbors=2, algorithm="ball_tree")
+    nbrs.fit(X_sub)
+
+    # Get min distances
+    distances, _ = nbrs.kneighbors(n_neighbors=1)
+    min_distance = distances[:, 0]
+
+    eps = 1e-13  # to get all neighbors of min distance
+    # For each i, collect all min-dist neighbors, remove self, pick one at random
+    nbr_i = np.empty(n, dtype=int)
+    for i in range(n):
+        # Query neighbors in the tight radius around the nearest neighbor distance
+        neighbors = nbrs.radius_neighbors(
+            X_sub[i, :].reshape(1, -1), min_distance[i] + eps, return_distance=False
+        )[0]
+        # Remove self index if present
+        neighbors = neighbors[neighbors != i]
+        nbr_i[i] = random_state.choice(neighbors)
+
+    return nbr_i
+
+
+def _nn_grouping_based(X_sub, random_state):
+    """Grouping-based NN selection with random tie-breaking.
+
+    Precondition: n_samples >= 2.
+    """
+    X_sub = np.asarray(X_sub)
+    n = X_sub.shape[0]
+
+    # 1) Group exactly identical rows
+    Xu, inv = np.unique(X_sub, axis=0, return_inverse=True)  # Xu: (m, p)
+    m = Xu.shape[0]
+
+    groups = [[] for _ in range(m)]
+    for i, g in enumerate(inv):
+        groups[g].append(i)
+    groups = [np.asarray(g, dtype=int) for g in groups]
+
+    nbr_i = np.empty(n, dtype=int)
+
+    for i in range(n):
+        gi = inv[i]
+        members = groups[gi]
+
+        # repeated data: choose another member of same group at random
+        if members.size >= 2:
+            choices = members[members != i]
+            nbr_i[i] = int(random_state.choice(choices))
+            continue
+
+        # per-query brute-force distances to all unique rows
+        diff = Xu - Xu[gi]  # (m, p)
+        d2 = (diff * diff).sum(axis=1)  # rowwise dot products
+        d2[gi] = np.inf  # exclude self
+
+        # Choose among all original indices whose (unique) row is at minimal distance
+        dmin = d2.min()
+        tied = np.flatnonzero(d2 == dmin)  # tied unique rows
+        candidates = np.concatenate([groups[u] for u in tied])
+        nbr_i[i] = int(random_state.choice(candidates))
+
+    return nbr_i
+
+
+def _Tn(X_sub, y_rank, random_state, *, nn_strategy="grouping"):
     """Compute :math:`T_n` following Fuchs (2024).
 
     The implementation uses the expression for :math:`T_n` given in
@@ -70,6 +141,8 @@ def _Tn(X_sub, y_rank, random_state):
         :func:`_rank_max`.
     random_state : numpy.random.RandomState
         Random number generator used to break nearest-neighbor ties.
+    nn_strategy : {"grouping", "radius"}, default="grouping"
+        Strategy used to select the nearest neighbor indices.
 
     Returns
     -------
@@ -78,25 +151,12 @@ def _Tn(X_sub, y_rank, random_state):
     """
     X_sub = np.asarray(X_sub)
     n = X_sub.shape[0]
-    # Fit NN on X_sub
-    nbrs = sklearn.neighbors.NearestNeighbors(n_neighbors=2, algorithm="ball_tree")
-    nbrs.fit(X_sub)
-
-    # Get min distances
-    distances, _ = nbrs.kneighbors(n_neighbors=1)
-    min_distance = distances[:, 0]
 
-    eps = 1e-13  # to get all neighbors of min distance
-    # For each i, collect all min-dist neighbors, remove self, pick one at random
-    nbr_i = np.empty(n, dtype=int)
-    for i in range(n):
-        # Query neighbors in the tight radius around the nearest neighbor distance
-        neighbors = nbrs.radius_neighbors(
-            X_sub[i, :].reshape(1, -1), min_distance[i] + eps, return_distance=False
-        )[0]
-        # Remove self index if present
-        neighbors = neighbors[neighbors != i]
-        nbr_i[i] = random_state.choice(neighbors)
+    if nn_strategy == "grouping":
+        nbr_i = _nn_grouping_based(X_sub, random_state)
+    else:
+        assert nn_strategy == "radius"
+        nbr_i = _nn_radius_based(X_sub, random_state)
 
     # Apply the formula (indices are 0-based; y_rank is 1-based)
     term1 = np.sum(np.abs(y_rank - y_rank[nbr_i]))
@@ -146,6 +206,9 @@ class FOCISelector(SelectorMixin, BaseEstimator):
         variance before computing nearest neighbors. If None, X is used as-is.
         Columns with zero variance are left unchanged.
 
+    nn_strategy : {"grouping", "radius"}, default="grouping"
+        Strategy used to select nearest neighbors for computing :math:`T_n`.
+
     random_state : int, RandomState instance or None, default=None
         Controls the random tie-breaking among nearest neighbors. Pass an int
         for reproducible results across multiple calls. If None, the global
@@ -180,15 +243,22 @@ class FOCISelector(SelectorMixin, BaseEstimator):
         "max_features": [None, Interval(Integral, 1, None, closed="left")],
         "min_delta": [None, Interval(Real, None, None, closed="neither")],
         "standardize": [None, StrOptions({"normalize"})],
+        "nn_strategy": [StrOptions({"grouping", "radius"})],
         "random_state": ["random_state"],
     }
 
     def __init__(
-        self, max_features=None, min_delta=0, standardize="normalize", random_state=None
+        self,
+        max_features=None,
+        min_delta=0,
+        standardize="normalize",
+        nn_strategy="grouping",
+        random_state=None,
     ):
         self.max_features = max_features
         self.min_delta = min_delta
         self.standardize = standardize
+        self.nn_strategy = nn_strategy
         self.random_state = random_state
 
     @_fit_context(prefer_skip_nested_validation=True)
@@ -249,7 +319,7 @@ class FOCISelector(SelectorMixin, BaseEstimator):
             for j in remaining:
                 sel_candidate = selected + [j]
                 X_sub = X[:, sel_candidate]
-                Tn_val = _Tn(X_sub, y_rank, random_state)
+                Tn_val = _Tn(X_sub, y_rank, random_state, nn_strategy=self.nn_strategy)
                 if Tn_val > best_Tn:
                     best_Tn = Tn_val
                     best_j = j

{pyfoci-0.2.2 → pyfoci-0.3.0}/pyFOCI/_version.py

@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
 commit_id: str | None
 __commit_id__: str | None
 
-__version__ = version = '0.2.2'
-__version_tuple__ = version_tuple = (0, 2, 2)
+__version__ = version = '0.3.0'
+__version_tuple__ = version_tuple = (0, 3, 0)
 
-__commit_id__ = commit_id = 'g62b06f720'
+__commit_id__ = commit_id = 'g3d446d680'

{pyfoci-0.2.2 → pyfoci-0.3.0}/pyFOCI/tests/test_foci.py

@@ -167,3 +167,38 @@ def test_random_state_int_reproducible():
         selector_1.Tn_path_,
         selector_2.Tn_path_,
     )
+
+
+def test_nn_strategy_grouping_and_radius_are_accepted_and_reproducible():
+    X_df, y = make_demo_data(n=200, p=10, seed=0)
+
+    for strategy in ("grouping", "radius"):
+        selector_1 = FOCISelector(
+            random_state=0, nn_strategy=strategy, min_delta=None, max_features=3
+        ).fit(X_df, y)
+        selector_2 = FOCISelector(
+            random_state=0, nn_strategy=strategy, min_delta=None, max_features=3
+        ).fit(X_df, y)
+
+        # Non-trivial selection (avoid early stopping selecting none)
+        assert selector_1.support_mask_.sum() > 2
+
+        np.testing.assert_array_equal(
+            selector_1.selected_indices_,
+            selector_2.selected_indices_,
+        )
+        assert_allclose(
+            selector_1.Tn_path_,
+            selector_2.Tn_path_,
+        )
+
+
+def test_nn_strategy_invalid_raises():
+    random_state = np.random.RandomState(0)
+    X = random_state.normal(size=(20, 3))
+    y = random_state.normal(size=20)
+
+    sel = FOCISelector(nn_strategy="invalid")
+    expected = "The 'nn_strategy' parameter of FOCISelector must be"
+    with pytest.raises(InvalidParameterError, match=re.escape(expected)):
+        sel.fit(X, y)

{pyfoci-0.2.2 → pyfoci-0.3.0/pyFOCI.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyFOCI
-Version: 0.2.2
+Version: 0.3.0
 Summary: A template for scikit-learn compatible packages.
 Author-email: Robert Pollak <robert.pollak@jku.at>
 License-Expression: BSD-3-Clause

{pyfoci-0.2.2 → pyfoci-0.3.0}/pyFOCI.egg-info/SOURCES.txt

@@ -41,6 +41,8 @@ pyFOCI.egg-info/PKG-INFO
 pyFOCI.egg-info/SOURCES.txt
 pyFOCI.egg-info/dependency_links.txt
 pyFOCI.egg-info/requires.txt
+pyFOCI.egg-info/scm_file_list.json
+pyFOCI.egg-info/scm_version.json
 pyFOCI.egg-info/top_level.txt
 pyFOCI/tests/__init__.py
 pyFOCI/tests/test_common.py

pyfoci-0.3.0/pyFOCI.egg-info/scm_file_list.json

@@ -0,0 +1,49 @@
+{
+  "files": [
+    ".coveragerc",
+    "CITATION.cff",
+    ".pre-commit-config.yaml",
+    "README.md",
+    "LICENSE",
+    ".codecov.yml",
+    "pyproject.toml",
+    "CHANGELOG.md",
+    "dev-readme.txt",
+    "pixi.lock",
+    ".gitignore",
+    "doc/Makefile",
+    "doc/quick_start.rst",
+    "doc/index.rst",
+    "doc/api.rst",
+    "doc/make.bat",
+    "doc/user_guide.rst",
+    "doc/conf.py",
+    "doc/_static/img/logo.png",
+    "doc/_static/img/index_getting_started.svg",
+    "doc/_static/img/index_examples.svg",
+    "doc/_static/img/index_user_guide.svg",
+    "doc/_static/img/index_api.svg",
+    "doc/_static/css/project-template.css",
+    "doc/_static/js/copybutton.js",
+    "doc/_templates/numpydoc_docstring.py",
+    "doc/_templates/function.rst",
+    "doc/_templates/sidebar-search-bs.html",
+    "doc/_templates/class.rst",
+    "pyFOCI/__init__.py",
+    "pyFOCI/_foci.py",
+    "pyFOCI/utils/__init__.py",
+    "pyFOCI/utils/discovery.py",
+    "pyFOCI/utils/tests/__init__.py",
+    "pyFOCI/utils/tests/test_discovery.py",
+    "pyFOCI/tests/__init__.py",
+    "pyFOCI/tests/test_foci.py",
+    "pyFOCI/tests/test_common.py",
+    "examples/plot_FOCISelector.py",
+    "examples/README.txt",
+    ".github/dependabot.yml",
+    ".github/workflows/release.yml",
+    ".github/workflows/python-app.yml",
+    ".github/workflows/lint.yml",
+    ".github/workflows/deploy-gh-pages.yml"
+  ]
+}

pyfoci-0.3.0/pyFOCI.egg-info/scm_version.json

@@ -0,0 +1,8 @@
+{
+  "tag": "0.3.0",
+  "distance": 0,
+  "node": "g3d446d6802cb189aeb6ca6c2e5623b56fb506933",
+  "dirty": false,
+  "branch": "HEAD",
+  "node_date": "2026-06-26"
+}