Merge pull request #52 from dynamicslab/check-optimizer-attributes

Check optimizer attributes

pysindy/optimizers/sindy_optimizer.py

@@ -32,8 +32,13 @@ class SINDyOptimizer(BaseEstimator):
     """
 
     def __init__(self, optimizer, unbias=True):
-        # TODO: add a check that optimizer has the necessary attributes
-        # and methods
+        if not hasattr(optimizer, "fit") or not callable(getattr(optimizer, "fit")):
+            raise AttributeError("optimizer does not have a callable fit method")
+        if not hasattr(optimizer, "predict") or not callable(
+            getattr(optimizer, "predict")
+        ):
+            raise AttributeError("optimizer does not have a callable predict method")
+
         self.optimizer = optimizer
         self.unbias = unbias
 
@@ -42,6 +47,8 @@ def fit(self, x, y):
             if not supports_multiple_targets(self.optimizer):
                 self.optimizer = _MultiTargetLinearRegressor(self.optimizer)
         self.optimizer.fit(x, y)
+        if not hasattr(self.optimizer, "coef_"):
+            raise AttributeError("optimizer has no attribute coef_")
         self.ind_ = np.abs(self.coef_) > 1e-14
 
         if self.unbias:
@@ -51,13 +58,18 @@ def fit(self, x, y):
 
     def _unbias(self, x, y):
         coef = np.zeros((y.shape[1], x.shape[1]))
+        if hasattr(self.optimizer, "fit_intercept"):
+            fit_intercept = self.optimizer.fit_intercept
+        else:
+            fit_intercept = False
+        if hasattr(self.optimizer, "normalize"):
+            normalize = self.optimizer.normalize
+        else:
+            normalize = False
         for i in range(self.ind_.shape[0]):
             if np.any(self.ind_[i]):
                 coef[i, self.ind_[i]] = (
-                    LinearRegression(
-                        fit_intercept=self.optimizer.fit_intercept,
-                        normalize=self.optimizer.normalize,
-                    )
+                    LinearRegression(fit_intercept=fit_intercept, normalize=normalize)
                     .fit(x[:, self.ind_[i]], y[:, i])
                     .coef_
                 )
@@ -82,7 +94,10 @@ def coef_(self):
 
     @property
     def intercept_(self):
-        return self.optimizer.intercept_
+        if hasattr(self.optimizer, "intercept_"):
+            return self.optimizer.intercept_
+        else:
+            return 0.0
 
     @property
     def complexity(self):

test/optimizers/test_optimizers.py

@@ -22,6 +22,26 @@ def fit(self, x, y):
         self.intercept_ = 0
         return self
 
+    def predict(self, x):
+        return x
+
+
+class DummyEmptyModel(BaseEstimator):
+    # Does not have fit or predict methods
+    def __init__(self):
+        self.fit_intercept = False
+        self.normalize = False
+
+
+class DummyModelNoCoef(BaseEstimator):
+    # Does not set the coef_ attribute
+    def fit(self, x, y):
+        self.intercept_ = 0
+        return self
+
+    def predict(self, x):
+        return x
+
 
 @pytest.mark.parametrize(
     "cls, support",
@@ -103,6 +123,18 @@ def test_bad_parameters(data_derivative_1d):
         SR3(max_iter=0)
 
 
+def test_bad_optimizers(data_derivative_1d):
+    x, x_dot = data_derivative_1d
+    x = x.reshape(-1, 1)
+
+    with pytest.raises(AttributeError):
+        opt = SINDyOptimizer(DummyEmptyModel())
+
+    with pytest.raises(AttributeError):
+        opt = SINDyOptimizer(DummyModelNoCoef())
+        opt.fit(x, x_dot)
+
+
 # The different capitalizations are intentional;
 # I want to make sure different versions are recognized
 @pytest.mark.parametrize("thresholder", ["L0", "l1", "CAD"])