test: add tests for dynamic learning rate decay

queens/stochastic_optimizers/learning_rate_decay.py

@@ -113,7 +113,7 @@ def __call__(self, learning_rate, params, gradient):
         return learning_rate
 
 
-class DynamicLearningRateDecay:
+class DynamicLearningRateDecay(LearningRateDecay):
     """Dynamic learning rate decay.
 
     Attributes:

tests/unit_tests/stochastic_optimizers/test_learning_rate_decay.py

@@ -0,0 +1,115 @@
+#
+# SPDX-License-Identifier: LGPL-3.0-or-later
+# Copyright (c) 2024, QUEENS contributors.
+#
+# This file is part of QUEENS.
+#
+# QUEENS is free software: you can redistribute it and/or modify it under the terms of the GNU
+# Lesser General Public License as published by the Free Software Foundation, either version 3 of
+# the License, or (at your option) any later version. QUEENS is distributed in the hope that it will
+# be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You
+# should have received a copy of the GNU Lesser General Public License along with QUEENS. If not,
+# see <https://www.gnu.org/licenses/>.
+#
+"""Unit tests for LearningRateDecay classes."""
+
+import numpy as np
+import pytest
+
+from queens.stochastic_optimizers.learning_rate_decay import (
+    DynamicLearningRateDecay,
+    LogLinearLearningRateDecay,
+    StepwiseLearningRateDecay,
+)
+
+
+def test_init_log_linear_learning_rate_decay():
+    """Test the init method of LogLinearLearningRateDecay class."""
+    slope = 0.7
+    learning_rate_decay = LogLinearLearningRateDecay(slope=slope)
+    assert learning_rate_decay.slope == slope
+    assert learning_rate_decay.iteration == 0
+
+
+def test_log_linear_learning_rate_decay():
+    """Test the call method of LogLinearLearningRateDecay class."""
+    learning_rate = 1.0
+    learning_rate_decay = LogLinearLearningRateDecay(slope=0.5)
+    num_iter = 101
+    for _ in range(num_iter):
+        learning_rate = learning_rate_decay(learning_rate, None, None)
+
+    assert learning_rate_decay.iteration == num_iter
+    np.testing.assert_array_almost_equal(learning_rate, 0.1)
+
+
+def test_init_stepwise_learning_rate_decay():
+    """Test the init method of StepwiseLearningRateDecay class."""
+    decay_factor = 0.1
+    decay_interval = 100
+    learning_rate_decay = StepwiseLearningRateDecay(
+        decay_factor=decay_factor, decay_interval=decay_interval
+    )
+    assert learning_rate_decay.decay_factor == decay_factor
+    assert learning_rate_decay.decay_interval == decay_interval
+    assert learning_rate_decay.iteration == 0
+
+
+def test_stepwise_learning_rate_decay():
+    """Test the call method of StepwiseLearningRateDecay class."""
+    learning_rate = 1.0
+    learning_rate_decay = StepwiseLearningRateDecay(decay_factor=0.1, decay_interval=10)
+    for _ in range(25):
+        learning_rate = learning_rate_decay(learning_rate, None, None)
+
+    assert learning_rate_decay.iteration == 3
+    np.testing.assert_array_almost_equal(learning_rate, 0.01)
+
+
+def test_init_dynamic_learning_rate_decay():
+    """Test the init method of DynamicLearningRateDecay class."""
+    alpha = 0.2
+    rho_min = 2.0
+    learning_rate_decay = DynamicLearningRateDecay(alpha=alpha, rho_min=rho_min)
+    assert learning_rate_decay.alpha == alpha
+    assert learning_rate_decay.rho_min == rho_min
+    assert learning_rate_decay.k_min == 2
+    assert learning_rate_decay.k == -1
+    assert learning_rate_decay.a == 0
+    assert learning_rate_decay.b == 0
+    assert learning_rate_decay.c == 0
+
+
+@pytest.mark.parametrize("alpha", [-1, 0, 1, 2])
+def test_init_dynamic_learning_rate_decay_invalid_alpha(alpha):
+    """Test the init method of DynamicLearningRateDecay class.
+
+    Test invalid values for alpha.
+    """
+    with pytest.raises(ValueError):
+        DynamicLearningRateDecay(alpha=alpha)
+
+
+@pytest.mark.parametrize("rho_min", [-1, 0])
+def test_init_dynamic_learning_rate_decay_invalid_rho_min(rho_min):
+    """Test the init method of DynamicLearningRateDecay class.
+
+    Test invalid values for rho_min.
+    """
+    with pytest.raises(ValueError):
+        DynamicLearningRateDecay(rho_min=rho_min)
+
+
+def test_dynamic_learning_rate_decay():
+    """Test the call method of LogLinearLearningRateDecay class."""
+    np.random.seed(1)
+    learning_rate = 1.0
+    learning_rate_decay = DynamicLearningRateDecay()
+    params = np.array([1.0, 2.0, 3.0])
+    num_iter = 101
+    for _ in range(num_iter):
+        learning_rate = learning_rate_decay(learning_rate, params, None)
+        params += np.random.randn(3) * 0.1
+
+    np.testing.assert_array_almost_equal(learning_rate, 0.01)