diff --git a/pyproject.toml b/pyproject.toml
index 509003c..70cb054 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -40,6 +40,7 @@ source = "https://github.com/JSS95/fluffyrocket"
 test = [
     "pytest",
     "aeon",
+    "scikit-learn",
 ]
 dev = [
     "flake8",
diff --git a/tests/test_learn.py b/tests/test_learn.py
new file mode 100644
index 0000000..e5e7cad
--- /dev/null
+++ b/tests/test_learn.py
@@ -0,0 +1,48 @@
+import torch
+from aeon.datasets import load_unit_test
+from sklearn.preprocessing import LabelEncoder
+
+from fluffyrocket import FluffyRocket
+
+
+def test_sharpness_learn():
+    """Test if sharpness parameter is learnable."""
+    # Load classification dataset.
+    X, y = load_unit_test()
+    X = X.astype("float32")
+
+    # Not-trainable model
+    fluffy = FluffyRocket(1.0, num_features=84, learnable=False, random_state=42)
+    fluffy.fit(X)
+    model = torch.nn.Sequential(
+        fluffy,
+        torch.nn.Linear(84, 2),
+    )
+    initial_sharpness = model[0].sharpness.clone()
+    trained_sharpness = train_by_one_step(model, X, y)[0].sharpness.clone()
+    assert initial_sharpness == trained_sharpness
+
+    # Trainable model
+    fluffy = FluffyRocket(1.0, num_features=84, learnable=True, random_state=42)
+    fluffy.fit(X)
+    model = torch.nn.Sequential(
+        fluffy,
+        torch.nn.Linear(84, 2),
+    )
+    initial_sharpness = model[0].sharpness.clone()
+    trained_sharpness = train_by_one_step(model, X, y)[0].sharpness.clone()
+    assert initial_sharpness != trained_sharpness
+
+
+def train_by_one_step(model, X, y):
+    X = torch.from_numpy(X)
+    y = torch.from_numpy(LabelEncoder().fit_transform(y))
+
+    optimizer = torch.optim.Adam(model.parameters())
+    loss_fn = torch.nn.CrossEntropyLoss()
+    optimizer.zero_grad()
+    output = model(X)
+    loss = loss_fn(output, y)
+    loss.backward()
+    optimizer.step()
+    return model