def test__acquire_higher_mean_equal_std(self):
# When the stds are equal but the mean s higher, it is selected
# Set-up
tuner = GPEi(tuple(), r_minimum=2)
tuner.y = np.array([0.5, 0.6, 0.7])
# Run
predictions = np.array([[0.8, 1], [0.9, 1]])
best = tuner._acquire(predictions)
# assert
assert best == 1
def test__acquire_best_ei_neq_best_y(self):
"""Best Expected Improvement does NOT correspond to the best prediction."""
# Set-up
tuner = GPEi(tuple(), r_minimum=2)
tuner.y = np.array([0.5, 0.6, 0.7])
# Run
predictions = np.array([[0.8, 2], [0.9, 1]])
best = tuner._acquire(predictions)
# assert
assert best == 0
def test__acquire_higher_mean_and_std(self):
# When both the mean and std of one point is higher, it is selected
# Set-up
tuner = GPEi(tuple(), r_minimum=2)
tuner.y = np.array([0.5, 0.6, 0.7])
# Run
predictions = np.array([[0.8, 1], [0.9, 2]])
best = tuner._acquire(predictions)
# assert
assert best == 1
def test_gpei(self):
X = [{'a': 1.1, 'b': 0.01, 'c': 3.5}, {'a': 4, 'b': 0.001, 'c': 6.2}]
y = [0.5, 0.6]
c1 = HyperParameter(ParamTypes.INT, [1, 5])
c2 = HyperParameter(ParamTypes.FLOAT_EXP, [0.0001, 0.1])
c3 = HyperParameter(ParamTypes.FLOAT, [2, 8])
tunables = [('a', c1), ('b', c2), ('c', c3)]
u = GPEi(tunables)
u.add(X, y)
for i in range(100):
proposed = u.propose()
self.assertTrue(proposed['a'] >= 1 and proposed['a'] <= 5)
self.assertTrue(proposed['b'] >= 0.0001 and proposed['b'] <= 0.1)
self.assertTrue(proposed['c'] >= 2 and proposed['c'] <= 8)
def test__acquire_best_ei_neq_best_y(self):
"""Best Expected Improvement does NOT correspond to the best prediction."""
# Set-up
tuner = GPEi(tuple(), r_minimum=2)
tuner.y = np.array([0.5, 0.6, 0.7])
# Run
predictions = np.array([
[0.8, 2],
[0.9, 1]
])
best = tuner._acquire(predictions)
# assert
assert best == 0
def test__acquire_possible_error(self):
"""Manually crafted case that seems to be an error in the formula:
The second prediction has a higher score and both have the same stdev.
However, the formula indicates that the first prediction is the best one.
"""
# Set-up
tuner = GPEi(tuple(), r_minimum=2)
tuner.y = np.array([0.5, 0.6, 0.7])
# Run
predictions = np.array([[0.8, 1], [0.9, 1]])
best = tuner._acquire(predictions)
# assert
assert best == 0
def test__acquire_lower_mean_higher_std(self):
# When the mean is higher but the std is lower, both outcomes are possible
# Create a situation with a much larger std to demonstrate.
# Set-up
tuner = GPEi(tuple(), r_minimum=2)
tuner.y = np.array([0.5, 0.6, 0.7])
# Run
predictions = np.array([
[0.9, 1],
[0.8, 100],
])
best = tuner._acquire(predictions)
# assert
assert best == 1
def test__acquire_possible_error(self):
"""Manually crafted case that seems to be an error in the formula:
The second prediction has a higher score and both have the same stdev.
However, the formula indicates that the first prediction is the best one.
"""
# Set-up
tuner = GPEi(tuple(), r_minimum=2)
tuner.y = np.array([0.5, 0.6, 0.7])
# Run
predictions = np.array([
[0.8, 1],
[0.9, 1]
])
best = tuner._acquire(predictions)
# assert
assert best == 0