Modified to enable the use of prior knowledge in the GPU-based DirectLiNGAM implementation

.github/workflows/tests.yml

@@ -1,6 +1,8 @@
 name: Tests
 
-on: [push]
+on:
+  push:
+  pull_request:
 
 jobs:
   build:
@@ -11,22 +13,69 @@ jobs:
         python-version: ["3.9", "3.10", "3.11", "3.12"]
 
     steps:
-      - uses: actions/checkout@v2
+      - uses: actions/checkout@v4
+
       - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v2
+        uses: actions/setup-python@v5
         with:
           python-version: ${{ matrix.python-version }}
+          cache: 'pip'
+
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
           pip install flake8 pytest pytest-cov
           if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+
       - name: Lint with flake8
         run: |
           # stop the build if there are Python syntax errors or undefined names
           flake8 lingam --count --select=E9,F63,F7,F82 --show-source --statistics
           # exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
           flake8 lingam --count --exit-zero --ignore=E203,E501,E741,C901 --max-line-length=127 --statistics
+
       - name: Test with pytest
         run: |
           pytest -v --cov=lingam --cov-report=term-missing
+
+  build-and-install:
+
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.9", "3.10", "3.11", "3.12"]
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: 'pip'
+
+      - name: Build sdist & wheel via pyproject.toml
+        run: |
+          python -m venv venv_b
+          source venv_b/bin/activate
+          python -m pip install --upgrade pip setuptools wheel build
+          python -m build
+          ls -la dist/
+          echo "List sdist contents:"
+          tar -tf dist/*.tar.gz | head -n 50
+
+      - name: Test install from wheel
+        run: |
+          python -m venv venv_w
+          source venv_w/bin/activate
+          pip install --no-cache-dir dist/*.whl
+          python -c "import lingam; print(lingam.__version__)"
+          pip check
+
+      - name: Test install from sdist
+        run: |
+          python -m venv venv_s
+          source venv_s/bin/activate
+          pip install --no-cache-dir dist/*.tar.gz
+          python -c "import lingam; print(lingam.__version__)"
+          pip check

lingam/causal_based_simulator.py

@@ -668,7 +668,7 @@ class CBSIUnobsCommonCauseLiNGAM(CBSILiNGAM):
     def _check_causal_graph(self, causal_graph, X):
         try:
             # fill nan with zeros
-            causal_graph = check_array(causal_graph, force_all_finite="allow-nan", copy=True)
+            causal_graph = check_array(causal_graph, ensure_all_finite="allow-nan", copy=True)
 
             n_features = X.shape[1]
             if causal_graph.shape != (n_features, n_features):

lingam/direct_lingam.py

@@ -252,13 +252,20 @@ def _search_causal_order_gpu(self, X, U):
             Returns the instance itself.
         mlist: causal ordering
         """
+        Uc, _ = self._search_candidate(U)
+        if len(Uc) == 1:
+            return Uc[0]
+
         cols = len(U)
         rows = len(X)
 
         arr = X[:, np.array(U)]
         from lingam_cuda import causal_order as causal_order_gpu
         mlist = causal_order_gpu(arr, rows, cols)
-        return U[np.argmax(mlist)]
+        M_list = np.asarray(mlist)
+        M_list = M_list[np.isin(U, Uc)]
+
+        return Uc[int(np.argmax(M_list))]
 
     def _mutual_information(self, x1, x2, param):
         """Calculate the mutual informations."""

lingam/tools/__init__.py

@@ -84,7 +84,7 @@ def bootstrap_with_imputation(
         Elements which are not NaN are the imputation values.
     """
     # check args
-    X = check_array(X, force_all_finite="allow-nan")
+    X = check_array(X, ensure_all_finite="allow-nan")
 
     n_sampling = check_scalar(n_sampling, "n_sampling", (numbers.Integral, np.integer), min_val=1)
 

lingam/utils/__init__.py

@@ -957,7 +957,7 @@ def evaluate_model_fit(adjacency_matrix, X, is_ordinal=None):
     """
 
     # check inputs
-    adj = check_array(adjacency_matrix, force_all_finite="allow-nan")
+    adj = check_array(adjacency_matrix, ensure_all_finite="allow-nan")
     if adj.ndim != 2 or (adj.shape[0] != adj.shape[1]):
         raise ValueError("adj must be an square matrix.")
 
@@ -1037,7 +1037,7 @@ def calculate_distance_from_root_nodes(adjacency_matrix, max_distance=None):
         adjacency_matrix,
         ensure_min_samples=2,
         ensure_min_features=2,
-        force_all_finite="allow-nan",
+        ensure_all_finite="allow-nan",
     )
     if adjacency_matrix.shape[0] != adjacency_matrix.shape[1]:
         raise ValueError("adjacency_matrix must be an square matrix.")
@@ -1103,7 +1103,7 @@ def calculate_total_effect(adjacency_matrix, from_index, to_index, is_continuous
         adjacency_matrix,
         ensure_min_samples=2,
         ensure_min_features=2,
-        force_all_finite="allow-nan",
+        ensure_all_finite="allow-nan",
     )
     if adjacency_matrix.shape[0] != adjacency_matrix.shape[1]:
         raise ValueError(

pyproject.toml

@@ -0,0 +1,49 @@
+[build-system]
+requires = ["setuptools>=61", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "lingam"
+description = "Python package for causal discovery based on LiNGAM."
+readme = "README.md"
+authors = [
+    { name = "Takashi Ikeuchi" },
+    { name = "Genya Haraoka" },
+    { name = "Mayumi Ide" },
+    { name = "Kouhei Nishikawa" },
+    { name = "Yan Zeng" },
+    { name = "Takashi Nicholas Maeda" },
+    { name = "Wataru Kurebayashi" },
+    { name = "Shohei Shimizu" }
+]
+license = { file = "LICENSE" }
+requires-python = ">=3.9"
+dependencies = [
+    "numpy",
+    "scipy",
+    "scikit-learn",
+    "graphviz",
+    "statsmodels",
+    "networkx",
+    "pandas",
+    "pygam",
+    "matplotlib",
+    "psy",
+    "semopy",
+    "autograd"
+]
+classifiers = [
+    "Programming Language :: Python :: 3",
+    "License :: OSI Approved :: MIT License",
+    "Operating System :: OS Independent"
+]
+dynamic = ["version"]
+
+[project.urls]
+Homepage = "https://github.com/cdt15/lingam"
+
+[tool.setuptools.packages.find]
+exclude = ["tests"]
+
+[tool.setuptools.dynamic]
+version = { attr = "lingam.__version__" }

tests/test_bottom_up_parce_lingam.py

@@ -102,9 +102,8 @@ def test_fit_invalid_data():
         raise AssertionError
 
     # Include non-numeric data
-    x0 = np.random.uniform(size=5)
-    x1 = np.array(['X', 'Y', 'X', 'Y', 'X'])
-    X = pd.DataFrame(np.array([x0, x1]).T, columns=['x0', 'x1'])
+    X = np.array([[1.2, 'X'],
+                  [3.4, 'Y']], dtype=object)
     try:
         model = BottomUpParceLiNGAM()
         model.fit(X)

tests/test_direct_lingam.py

@@ -56,9 +56,8 @@ def test_fit_invalid_data():
         raise AssertionError
 
     # Include non-numeric data
-    x0 = np.random.uniform(size=5)
-    x1 = np.array(['X', 'Y', 'X', 'Y', 'X'])
-    X = pd.DataFrame(np.array([x0, x1]).T, columns=['x0', 'x1'])
+    X = np.array([[1.2, 'X'],
+                  [3.4, 'Y']], dtype=object)
     try:
         model = DirectLiNGAM()
         model.fit(X)

tests/test_high_dim_direct_lingam.py

@@ -73,9 +73,8 @@ def test_fit_invalid_data():
         raise AssertionError
 
     # Include non-numeric data
-    x0 = np.random.uniform(size=5)
-    x1 = np.array(['X', 'Y', 'X', 'Y', 'X'])
-    X = pd.DataFrame(np.array([x0, x1]).T, columns=['x0', 'x1'])
+    X = np.array([[1.2, 'X'],
+                  [3.4, 'Y']], dtype=object)
     try:
         model = HighDimDirectLiNGAM()
         model.fit(X)

tests/test_rcd.py

@@ -107,9 +107,8 @@ def test_fit_invalid_data():
         raise AssertionError
 
     # Include non-numeric data
-    x0 = np.random.uniform(size=5)
-    x1 = np.array(['X', 'Y', 'X', 'Y', 'X'])
-    X = pd.DataFrame(np.array([x0, x1]).T, columns=['x0', 'x1'])
+    X = np.array([[1.2, 'X'],
+                  [3.4, 'Y']], dtype=object)
     try:
         model = RCD()
         model.fit(X)