Merge remote-tracking branch 'qmlcode/develop' into qmlcode#90

Author: larsbratholm

docs/source/qml_examples/examples.ipynb

@@ -39,8 +39,9 @@
 
 ## ------------- ** Setting up the estimator ** ---------------
 
-estimator = ARMP(iterations=10, representation='acsf', representation_params={"radial_rs": np.arange(0, 10, 1), "angular_rs": np.arange(0.5, 10.5, 1),
-"theta_s": np.arange(0, 5, 1)}, tensorboard=False)
+acsf_params = {"nRs2": 5, "nRs3": 5, "nTs": 5, "rcut": 5, "acut": 5, "zeta": 220.127, "eta": 30.8065}
+estimator = ARMP(iterations=5000, representation_name='acsf', representation_params=acsf_params, tensorboard=False,
+                 learning_rate=0.075, l1_reg=0.0, l2_reg=0.0)
 
 estimator.generate_compounds(filenames)
 estimator.set_properties(energies)

examples/ARMP_1.py

@@ -39,7 +39,7 @@
 
 ## ------------- ** Setting up the estimator ** ---------------
 
-estimator = ARMP(iterations=100, l2_reg=0.0)
+estimator = ARMP(iterations=3000, learning_rate=0.075, l1_reg=0.0, l2_reg=0.0, tensorboard=True, store_frequency=50)
 
 estimator.set_representations(representations=descriptor)
 estimator.set_classes(zs)

examples/ARMP_2.py

@@ -39,7 +39,7 @@
 
 ## ------------- ** Setting up the estimator ** ---------------
 
-estimator = ARMP(iterations=150, l2_reg=0.0, learning_rate=0.005, hidden_layer_sizes=(40, 20, 10))
+estimator = ARMP(iterations=3000, learning_rate=0.075, l1_reg=0.0, l2_reg=0.0, tensorboard=True, store_frequency=50)
 
 ##  ------------- ** Fitting to the data ** ---------------
 

examples/ARMP_3.py

@@ -0,0 +1,41 @@
+"""
+This example shows how to use ARMP to overfit 100 data-points for the QM7 data set. It uses the Atom Centred Symmetry
+functions as the representation.
+
+This example takes about 3.5 min to run on a mac.
+"""
+
+from qml.aglaia.aglaia import ARMP
+import glob
+import numpy as np
+import matplotlib.pyplot as plt
+from sklearn import model_selection as modsel
+
+filenames = sorted(glob.glob("../test/qm7/*.xyz"))
+energies = np.loadtxt("../test/data/hof_qm7.txt", usecols=[1])
+n_samples = len(filenames)
+print("%i files were loaded." % (n_samples))
+
+acsf_params = {"nRs2": 5, "nRs3": 5, "nTs": 5, "rcut": 5, "acut": 5, "zeta": 220.127, "eta": 30.8065}
+estimator = ARMP(iterations=6000, representation_name='acsf', representation_params=acsf_params, l1_reg=0.0, l2_reg=0.0,
+                 scoring_function="rmse", tensorboard=False, store_frequency=10, learning_rate=0.075)
+
+estimator.set_properties(energies[:100])
+estimator.generate_compounds(filenames[:100])
+estimator.generate_representation(method="fortran")
+print("The shape of the representation is: %s" % (str(estimator.representation.shape)))
+
+idx = list(range(100))
+
+idx_train, idx_test = modsel.train_test_split(idx, test_size=0, random_state=42, shuffle=True)
+
+estimator.fit(idx_train)
+
+score = estimator.score(idx_train)
+print("The RMSE is %s kcal/mol." % (str(score)))
+
+ene_pred = estimator.predict(idx_train)
+
+# Plotting the predictions against the true values
+plt.scatter(energies[idx_train], ene_pred)
+plt.show()

examples/ARMP_qm7.py

@@ -41,7 +41,7 @@
 
 ## ------------- ** Setting up the estimator ** ---------------
 
-estimator = MRMP(representation='slatm', representation_params={'slatm_dgrid2': 0.06, 'slatm_dgrid1': 0.06})
+estimator = MRMP(representation_name='slatm', representation_params={'slatm_dgrid2': 0.06, 'slatm_dgrid1': 0.06})
 
 estimator.generate_compounds(filenames[:100])
 estimator.set_properties(energies[:100])

examples/MRMP_1.py

@@ -213,6 +213,64 @@ def pipelines():
     print("*** End pipelines examples ***")
     print()
 
+def pipelines_2():
+    """
+    Scikit learn pipeline with a molecular neural network
+    """
+
+    print("\n *** Begin pipelines example with molecular Neural Network ***")
+
+    data = qmlearn.Data("../test/qm7/*.xyz")
+    energies = np.loadtxt("../test/data/hof_qm7.txt", usecols=1)
+    data.set_energies(energies)
+
+    # Create model
+    model = sklearn.pipeline.make_pipeline(
+        qmlearn.preprocessing.AtomScaler(data),
+        qmlearn.representations.CoulombMatrix(),
+        qmlearn.models.NeuralNetwork(iterations=500, batch_size=50, learning_rate=0.005),
+    )
+
+    indices = np.arange(1000)
+    np.random.shuffle(indices)
+
+    model.fit(indices[:100])
+
+    # Score on the TRAINING set, since you won't get good predictions in 500 iterations
+    scores = model.score(indices[:100])
+    print("Negative MAE:", scores)
+
+    print("*** End pipelines example with molecular Neural Network *** \n")
+
+def pipelines_3():
+    """
+    Scikit learn pipeline with an atomic neural network
+    """
+
+    print("\n *** Begin pipelines example with atomic Neural Network ***")
+
+    data = qmlearn.Data("../test/qm7/*.xyz")
+    energies = np.loadtxt("../test/data/hof_qm7.txt", usecols=1)
+    data.set_energies(energies)
+
+    # Create model
+    model = sklearn.pipeline.make_pipeline(
+        qmlearn.preprocessing.AtomScaler(data),
+        qmlearn.representations.AtomCenteredSymmetryFunctions(),
+        qmlearn.models.NeuralNetwork(iterations=500, batch_size=50, learning_rate=0.005),
+    )
+
+    indices = np.arange(1000)
+    np.random.shuffle(indices)
+
+    model.fit(indices[:100])
+
+    # Score on the TRAINING set, since you won't get good predictions in 500 iterations
+    scores = model.score(indices[:100])
+    print("Negative MAE:", scores)
+
+    print("*** End pipelines example with atomic Neural Network *** \n")
+
 def cross_validation():
     """
     Doing cross validation with qmlearn
@@ -285,3 +343,5 @@ def cross_validation():
     models()
     pipelines()
     cross_validation()
+    pipelines_2()
+    pipelines_3()