def test_qsvm_binary_directly_statevector(self):
""" QSVM Binary Directly Statevector test """
ref_kernel_testing = np. array([[0.1443953, 0.18170069, 0.47479649, 0.14691763],
[0.33041779, 0.37663733, 0.02115561, 0.16106199]])
ref_support_vectors = np.array([[2.95309709, 2.51327412], [3.14159265, 4.08407045],
[4.08407045, 2.26194671], [4.46106157, 2.38761042]])
backend = BasicAer.get_backend('statevector_simulator')
num_qubits = 2
feature_map = SecondOrderExpansion(feature_dimension=num_qubits,
depth=2,
entangler_map=[[0, 1]])
svm = QSVM(feature_map, self.training_data, self.testing_data, None)
quantum_instance = QuantumInstance(backend, seed_transpiler=self.random_seed,
seed_simulator=self.random_seed)
file_path = self.get_resource_path('qsvm_test.npz')
try:
result = svm.run(quantum_instance)
ori_alphas = result['svm']['alphas']
np.testing.assert_array_almost_equal(
result['kernel_matrix_testing'], ref_kernel_testing, decimal=4)
self.assertEqual(len(result['svm']['support_vectors']), 4)
np.testing.assert_array_almost_equal(
result['svm']['support_vectors'], ref_support_vectors, decimal=4)
self.assertEqual(result['testing_accuracy'], 0.5)
svm.save_model(file_path)
self.assertTrue(os.path.exists(file_path))
loaded_svm = QSVM(feature_map)
loaded_svm.load_model(file_path)
np.testing.assert_array_almost_equal(
loaded_svm.ret['svm']['support_vectors'], ref_support_vectors, decimal=4)
np.testing.assert_array_almost_equal(
loaded_svm.ret['svm']['alphas'], ori_alphas, decimal=4)
loaded_test_acc = loaded_svm.test(svm.test_dataset[0],
svm.test_dataset[1],
quantum_instance)
self.assertEqual(result['testing_accuracy'], loaded_test_acc)
np.testing.assert_array_almost_equal(
loaded_svm.ret['kernel_matrix_testing'], ref_kernel_testing, decimal=4)
except NameError as ex:
self.skipTest(str(ex))
finally:
if os.path.exists(file_path):
try:
os.remove(file_path)
except Exception: # pylint: disable=broad-except
pass
def test_binary_directly_statevector(self):
"""Test QSVM on binary classification on BasicAer's statevector simulator.
Also tests saving and loading models."""
data_preparation = self.data_preparation
svm = QSVM(data_preparation, self.training_data, self.testing_data,
None)
file_path = self.get_resource_path('qsvm_test.npz')
try:
result = svm.run(self.statevector_simulator)
ori_alphas = result['svm']['alphas']
np.testing.assert_array_almost_equal(
result['kernel_matrix_testing'],
self.ref_kernel_testing['statevector'],
decimal=4)
self.assertEqual(len(result['svm']['support_vectors']), 4)
np.testing.assert_array_almost_equal(
result['svm']['support_vectors'],
self.ref_support_vectors,
decimal=4)
self.assertEqual(result['testing_accuracy'], 0.5)
svm.save_model(file_path)
self.assertTrue(os.path.exists(file_path))
loaded_svm = QSVM(data_preparation)
loaded_svm.load_model(file_path)
np.testing.assert_array_almost_equal(
loaded_svm.ret['svm']['support_vectors'],
self.ref_support_vectors,
decimal=4)
np.testing.assert_array_almost_equal(
loaded_svm.ret['svm']['alphas'], ori_alphas, decimal=4)
loaded_test_acc = loaded_svm.test(svm.test_dataset[0],
svm.test_dataset[1],
self.statevector_simulator)
self.assertEqual(result['testing_accuracy'], loaded_test_acc)
np.testing.assert_array_almost_equal(
loaded_svm.ret['kernel_matrix_testing'],
self.ref_kernel_testing['statevector'],
decimal=4)
except MissingOptionalLibraryError as ex:
self.skipTest(str(ex))
finally:
if os.path.exists(file_path):
try:
os.remove(file_path)
except Exception: # pylint: disable=broad-except
pass
def compareMethods(class1, class2, class3 = None, backend=BasicAer.get_backend('qasm_simulator'), name = "",
include_unscaled=False, include_QSVM = True, include_VQC = True, feature_dimension = 2, gamma = 'auto', C = 1.0):
#Define header and chart data
data = []
header = ["Algorithm", "Backend", "Time", "Accuracy", "Only one Class Predicted?"]
data.append(header)
#Split data into train and test
class1_train, class1_test = train_test_split(class1, test_size=0.33, random_state=42)
class2_train, class2_test = train_test_split(class2, test_size=0.33, random_state=42)
feature_dim = feature_dimension
if class3 is not None:
class3_train, class3_test = train_test_split(class3, test_size=0.33, random_state=42)
#Get input data for quantum
training_data = {'A': np.asarray(class1_train), 'B': np.asarray(class2_train)}
test_data = {'A': np.asarray(class1_test), 'B': np.asarray(class2_test)}
total_array = np.concatenate((test_data['A'], test_data['B']))
if class3 is not None:
training_data["C"] = class3_train
test_data["C"] = class3_test
total_array = np.concatenate((total_array, test_data['C']))
#Get input data for classical
X_train, x_test, Y_train, y_test = convertFromQS(training_data, test_data)
#Classical SVM, linear kernel (scaled and unscaled)
if include_unscaled:
start = time.time()
clf = svm.SVC(kernel='linear') # Linear Kernel
model = clf.fit(X_train, Y_train)
y_pred = clf.predict(x_test)
end = time.time()
data.append(["SVM, Linear Kernel", "Local Processor", round(end-start), str(round(100*metrics.accuracy_score(y_test, y_pred), 2)),checkAllSame(y_pred)])
start = time.time()
scaler = StandardScaler()
X_train_std = scaler.fit_transform(X_train)
x_test_std = scaler.fit_transform(x_test)
clf = svm.SVC(kernel='linear') # Linear Kernel
model = clf.fit(X_train_std, Y_train)
y_pred = clf.predict(x_test_std)
end = time.time()
data.append(["SVM, Linear Kernel, scaled", "Local Processor", round(end-start), str(round(100*metrics.accuracy_score(y_test, y_pred), 2)),checkAllSame(y_pred)])
#Classical SVM, rbf kernel (scaled and unscaled)
if include_unscaled:
start = time.time()
clf = svm.SVC(C=C, kernel='rbf', gamma = gamma) # rbf Kernel
model = clf.fit(X_train, Y_train)
y_pred = clf.predict(x_test)
end = time.time()
data.append(["SVM, RBF Kernel", "Local Processor", round(end-start), str(round(100*metrics.accuracy_score(y_test, y_pred), 2)),checkAllSame(y_pred)])
start = time.time()
scaler = StandardScaler()
X_train_std = scaler.fit_transform(X_train)
x_test_std = scaler.fit_transform(x_test)
clf = svm.SVC(C=C, kernel='rbf', gamma = gamma) # rbf Kernel
model = clf.fit(X_train_std, Y_train)
y_pred = clf.predict(x_test_std)
end = time.time()
data.append(["SVM, RBF Kernel, scaled", "Local Processor", round(end-start), str(round(100*metrics.accuracy_score(y_test, y_pred), 2)),checkAllSame(y_pred)])
#QSVM run
if include_QSVM:
start = time.time()
feature_map = ZZFeatureMap(feature_dimension=feature_dim, reps=2, entanglement='linear')
if class3 is None:
qsvm = QSVM(feature_map, training_data, test_data, total_array)
else:
qsvm = QSVM(feature_map, training_data, test_data, total_array, multiclass_extension=AllPairs())
quantum_instance = QuantumInstance(backend, shots=1024, seed_simulator=10598, seed_transpiler=10598)
resultSVM = qsvm.run(quantum_instance)
end = time.time()
QSVM_Summary = ["QSVM", backend.name(), round(end-start), str(round(100*resultSVM['testing_accuracy'], 2)), checkAllSame(resultSVM['predicted_classes'])]
data.append(QSVM_Summary)
path = 'C:\\Users\\admin\\Desktop\\QQML\\Code\\Saved_SVMs\\' + name + "_" + backend.name() + "_QSVM"
if class3 is None: #Bug in package prevents saving Multiclass svms. Will find workaround or submit bug report if time.
qsvm.save_model(path)
#VQC run
if include_VQC:
start = time.time()
optimizer = SPSA(max_trials=100, c0=4.0, skip_calibration=True)
optimizer.set_options(save_steps=1)
feature_map = ZZFeatureMap(feature_dimension=feature_dim, reps=2)
var_form = TwoLocal(feature_dim, ['ry', 'rz'], 'cz', reps=3)
vqc = VQC(optimizer, feature_map, var_form, training_data, test_data, total_array)
quantum_instance = QuantumInstance(backend, shots=1024, seed_simulator=10589, seed_transpiler=10598)
resultVQC = vqc.run(quantum_instance)
end = time.time()
VQC_Summary = ["VQC", backend.name(), round(end-start), str(round(100*resultVQC['testing_accuracy'], 2)), checkAllSame(resultVQC['predicted_classes'])]
data.append(VQC_Summary)
path = 'C:\\Users\\admin\\Desktop\\QQML\\Code\\Saved_SVMs\\' + name + "_" + backend.name() + "_VQC"
vqc.save_model(path)
display(HTML(tabulate.tabulate(data, tablefmt='html')))
return data