modAL.acquisition functions refactored

Author: cosmic-cortex

examples/bayesian_optimization.py

@@ -4,7 +4,7 @@
 from sklearn.gaussian_process import GaussianProcessRegressor
 from sklearn.gaussian_process.kernels import Matern
 from modAL.models import BayesianOptimizer
-from modAL.acquisition import PI, EI, UCB, max_PI, max_EI, max_UCB
+from modAL.acquisition import optimizer_PI, optimizer_EI, optimizer_UCB, max_PI, max_EI, max_UCB
 
 
 # generating the data
@@ -24,12 +24,12 @@
 kernel = Matern(length_scale=1.0)
 
 tr = 0.1
-PI_tr = partial(PI, tradeoff=tr)
+PI_tr = partial(optimizer_PI, tradeoff=tr)
 PI_tr.__name__ = 'PI, tradeoff = %1.1f' % tr
 max_PI_tr = partial(max_PI, tradeoff=tr)
 
 acquisitions = zip(
-    [PI_tr, EI, UCB],
+    [PI_tr, optimizer_EI, optimizer_UCB],
     [max_PI_tr, max_EI, max_UCB],
 )
 

modAL/__init__.py

@@ -1,5 +1,5 @@
 from .models import ActiveLearner, Committee, CommitteeRegressor
-from .acquisition import PI, EI, UCB, max_PI, max_EI, max_UCB
+from .acquisition import PI, EI, UCB, optimizer_PI, optimizer_EI, optimizer_UCB, max_PI, max_EI, max_UCB
 from .uncertainty import classifier_uncertainty, classifier_margin, classifier_entropy, \
                          uncertainty_sampling, margin_sampling, entropy_sampling
 from .disagreement import vote_entropy, consensus_entropy, KL_max_disagreement, \
@@ -8,7 +8,7 @@
 
 __all__ = [
     'ActiveLearner', 'Committee', 'CommitteeRegressor',
-    'PI', 'EI', 'UCB', 'max_PI', 'max_EI', 'max_UCB',
+    'PI', 'EI', 'UCB', 'optimizer_PI', 'optimizer_EI', 'optimizer_UCB', 'max_PI', 'max_EI', 'max_UCB',
     'classifier_uncertainty', 'classifier_margin', 'classifier_entropy',
     'uncertainty_sampling', 'margin_sampling', 'entropy_sampling',
     'vote_entropy', 'consensus_entropy', 'KL_max_disagreement',

modAL/acquisition.py

@@ -12,7 +12,20 @@
 from modAL.utils.selection import multi_argmax
 
 
-def PI(optimizer, X, tradeoff=0):
+def PI(mean, std, max_val, tradeoff):
+    return ndtr((mean - max_val - tradeoff)/std)
+
+
+def EI(mean, std, max_val, tradeoff):
+    z = (mean - max_val - tradeoff) / std
+    return (mean - max_val - tradeoff)*ndtr(z) + std*norm.pdf(z)
+
+
+def UCB(mean, std, beta):
+    return mean + beta*std
+
+
+def optimizer_PI(optimizer, X, tradeoff=0):
     """
     Probability of improvement acquisition function for Bayesian optimization.
 
@@ -35,10 +48,10 @@ def PI(optimizer, X, tradeoff=0):
     mean, std = optimizer.predict(X, return_std=True)
     std = std.reshape(-1, 1)
 
-    return ndtr((mean - optimizer.max_val - tradeoff)/std)
+    return PI(mean, std, optimizer.max_val, tradeoff)
 
 
-def EI(optimizer, X, tradeoff=0):
+def optimizer_EI(optimizer, X, tradeoff=0):
     """
     Expected improvement acquisition function for Bayesian optimization.
 
@@ -60,12 +73,11 @@ def EI(optimizer, X, tradeoff=0):
     """
     mean, std = optimizer.predict(X, return_std=True)
     std = std.reshape(-1, 1)
-    z = (mean - optimizer.max_val - tradeoff)/std
 
-    return (mean - optimizer.max_val - tradeoff)*ndtr(z) + std*norm.pdf(z)
+    return EI(mean, std, optimizer.max_val, tradeoff)
 
 
-def UCB(optimizer, X, beta=1):
+def optimizer_UCB(optimizer, X, beta=1):
     """
     Upper confidence bound acquisition function for Bayesian optimization.
 
@@ -88,7 +100,7 @@ def UCB(optimizer, X, beta=1):
     mean, std = optimizer.predict(X, return_std=True)
     std = std.reshape(-1, 1)
 
-    return mean + beta*std
+    return UCB(mean, std, beta)
 
 
 def max_PI(optimizer, X, tradeoff=0, n_instances=1):
@@ -117,7 +129,7 @@ def max_PI(optimizer, X, tradeoff=0, n_instances=1):
     X[query_idx]: numpy.ndarray of shape (n_instances, n_features)
         The instances from X chosen to be labelled.
     """
-    pi = PI(optimizer, X, tradeoff=tradeoff)
+    pi = optimizer_PI(optimizer, X, tradeoff=tradeoff)
     query_idx = multi_argmax(pi, n_instances=n_instances)
 
     return query_idx, X[query_idx]
@@ -149,7 +161,7 @@ def max_EI(optimizer, X, tradeoff=0, n_instances=1):
     X[query_idx]: numpy.ndarray of shape (n_instances, n_features)
         The instances from X chosen to be labelled.
     """
-    ei = EI(optimizer, X, tradeoff=tradeoff)
+    ei = optimizer_EI(optimizer, X, tradeoff=tradeoff)
     query_idx = multi_argmax(ei, n_instances=n_instances)
 
     return query_idx, X[query_idx]
@@ -182,7 +194,7 @@ def max_UCB(optimizer, X, beta=1, n_instances=1):
     X[query_idx]: numpy.ndarray of shape (n_instances, n_features)
         The instances from X chosen to be labelled.
     """
-    ucb = UCB(optimizer, X, beta=beta)
+    ucb = optimizer_UCB(optimizer, X, beta=beta)
     query_idx = multi_argmax(ucb, n_instances=n_instances)
 
     return query_idx, X[query_idx]

modAL/models.py

@@ -410,7 +410,7 @@ class BayesianOptimizer(BaseLearner):
     >>> from sklearn.gaussian_process import GaussianProcessRegressor
     >>> from sklearn.gaussian_process.kernels import Matern
     >>> from modAL.models import BayesianOptimizer
-    >>> from modAL.acquisition import PI, EI, UCB, max_PI, max_EI, max_UCB
+    >>> from modAL.acquisition import optimizer_PI, optimizer_EI, optimizer_UCB, max_PI, max_EI, max_UCB
     >>>
     >>> # generating the data
     >>> X = np.linspace(0, 20, 1000).reshape(-1, 1)
@@ -423,12 +423,12 @@ class BayesianOptimizer(BaseLearner):
     >>> kernel = Matern(length_scale=1.0)
     >>>
     >>> tr = 0.1
-    >>> PI_tr = partial(PI, tradeoff=tr)
+    >>> PI_tr = partial(optimizer_PI, tradeoff=tr)
     >>> PI_tr.__name__ = 'PI, tradeoff = %1.1f' % tr
     >>> max_PI_tr = partial(max_PI, tradeoff=tr)
     >>>
     >>> acquisitions = zip(
-    ...     [PI_tr, EI, UCB],
+    ...     [PI_tr, optimizer_EI, optimizer_UCB],
     ...     [max_PI_tr, max_EI, max_UCB],
     ... )
     >>>

tests/core_tests.py

@@ -126,7 +126,19 @@ def test_make_query_strategy(self):
 
 
 class TestAcquisitionFunctions(unittest.TestCase):
-    def test_PI(self):
+    def test_acquisition_functions(self):
+        for n_samples in range(1, 100):
+            mean, std = np.random.rand(100, 1), np.random.rand(100, 1)
+            modAL.acquisition.PI(mean, std, 0, 0)
+            modAL.acquisition.EI(mean, std, 0, 0)
+            modAL.acquisition.UCB(mean, std, 0)
+
+            mean, std = np.random.rand(100, ), np.random.rand(100, )
+            modAL.acquisition.PI(mean, std, 0, 0)
+            modAL.acquisition.EI(mean, std, 0, 0)
+            modAL.acquisition.UCB(mean, std, 0)
+
+    def test_optimizer_PI(self):
         for n_samples in range(1, 100):
             mean = np.random.rand(n_samples, 1)
             std = np.random.rand(n_samples, 1)
@@ -142,10 +154,10 @@ def test_PI(self):
 
             np.testing.assert_almost_equal(
                 ndtr((mean - max_val - tradeoff)/std),
-                modAL.acquisition.PI(optimizer, np.random.rand(n_samples, 2), tradeoff)
+                modAL.acquisition.optimizer_PI(optimizer, np.random.rand(n_samples, 2), tradeoff)
             )
 
-    def test_EI(self):
+    def test_optimizer_EI(self):
         for n_samples in range(1, 100):
             mean = np.random.rand(n_samples, 1)
             std = np.random.rand(n_samples, 1)
@@ -164,10 +176,10 @@ def test_EI(self):
 
             np.testing.assert_almost_equal(
                 true_EI,
-                modAL.acquisition.EI(optimizer, np.random.rand(n_samples, 2), tradeoff)
+                modAL.acquisition.optimizer_EI(optimizer, np.random.rand(n_samples, 2), tradeoff)
             )
 
-    def test_UCB(self):
+    def test_optimizer_UCB(self):
         for n_samples in range(1, 100):
             mean = np.random.rand(n_samples, 1)
             std = np.random.rand(n_samples, 1)
@@ -183,7 +195,7 @@ def test_UCB(self):
 
             np.testing.assert_almost_equal(
                 true_UCB,
-                modAL.acquisition.UCB(optimizer, np.random.rand(n_samples, 2), beta)
+                modAL.acquisition.optimizer_UCB(optimizer, np.random.rand(n_samples, 2), beta)
             )
 
     def test_selection(self):

tests/example_tests/bayesian_optimization.py

@@ -3,7 +3,7 @@
 from sklearn.gaussian_process import GaussianProcessRegressor
 from sklearn.gaussian_process.kernels import Matern
 from modAL.models import BayesianOptimizer
-from modAL.acquisition import PI, EI, UCB, max_PI, max_EI, max_UCB
+from modAL.acquisition import optimizer_PI, optimizer_EI, optimizer_UCB, max_PI, max_EI, max_UCB
 
 
 # generating the data
@@ -17,12 +17,12 @@
 kernel = Matern(length_scale=1.0)
 
 tr = 0.1
-PI_tr = partial(PI, tradeoff=tr)
+PI_tr = partial(optimizer_PI, tradeoff=tr)
 PI_tr.__name__ = 'PI, tradeoff = %1.1f' % tr
 max_PI_tr = partial(max_PI, tradeoff=tr)
 
 acquisitions = zip(
-    [PI_tr, EI, UCB],
+    [PI_tr, optimizer_EI, optimizer_UCB],
     [max_PI_tr, max_EI, max_UCB],
 )