add TestSurrogateNamespace

Author: SimonBlanke

tests/full/custom_test_function/__init__.py

@@ -0,0 +1,82 @@
+# Author: Simon Blanke
+# Email: simon.blanke@yahoo.com
+# License: MIT License
+
+"""Tests for CustomTestFunction surrogate namespace (requires scikit-learn)."""
+
+import numpy as np
+import pytest
+
+from surfaces.custom_test_function import CustomTestFunction
+
+
+@pytest.fixture
+def sphere_func():
+    """Simple sphere function for testing."""
+
+    def sphere(params):
+        return sum(v**2 for v in params.values())
+
+    return CustomTestFunction(
+        objective_fn=sphere,
+        search_space={"x": (-5, 5), "y": (-5, 5)},
+    )
+
+
+@pytest.fixture
+def sphere_func_with_data(sphere_func):
+    """Sphere function with evaluation data."""
+    np.random.seed(42)
+    for _ in range(50):
+        x, y = np.random.uniform(-5, 5, 2)
+        sphere_func({"x": x, "y": y})
+    return sphere_func
+
+
+class TestSurrogateNamespace:
+    """Test surrogate namespace methods."""
+
+    def test_fit_random_forest(self, sphere_func_with_data):
+        """Test Random Forest surrogate fitting."""
+        sphere_func_with_data.surrogate.fit(method="random_forest")
+
+        assert sphere_func_with_data.surrogate.is_fitted
+        assert sphere_func_with_data.surrogate.method == "random_forest"
+
+    def test_predict(self, sphere_func_with_data):
+        """Test surrogate prediction."""
+        sphere_func_with_data.surrogate.fit(method="random_forest")
+
+        # Predict at origin (should be near 0)
+        pred = sphere_func_with_data.surrogate.predict({"x": 0, "y": 0})
+        assert pred < 5  # Should be small
+
+    def test_predict_array(self, sphere_func_with_data):
+        """Test surrogate prediction with array input."""
+        sphere_func_with_data.surrogate.fit(method="random_forest")
+
+        X = np.array([[0, 0], [1, 1], [2, 2]])
+        preds = sphere_func_with_data.surrogate.predict(X)
+
+        assert len(preds) == 3
+
+    def test_suggest_next(self, sphere_func_with_data):
+        """Test next point suggestion."""
+        sphere_func_with_data.surrogate.fit(method="random_forest")
+
+        suggestions = sphere_func_with_data.surrogate.suggest_next(n_suggestions=3)
+
+        assert len(suggestions) == 3
+        assert all("x" in s and "y" in s for s in suggestions)
+
+    def test_score(self, sphere_func_with_data):
+        """Test surrogate R^2 score."""
+        sphere_func_with_data.surrogate.fit(method="random_forest")
+
+        score = sphere_func_with_data.surrogate.score()
+        assert 0 < score <= 1  # R^2 should be positive for good fit
+
+    def test_not_fitted_error(self, sphere_func_with_data):
+        """Test error when predicting without fitting."""
+        with pytest.raises(RuntimeError, match="No surrogate model fitted"):
+            sphere_func_with_data.surrogate.predict({"x": 0, "y": 0})