def run_exp(
method='qrac',
epochs=300,
positive_factor=1 / 3,
depth=4,
seed=10598,
reg=0.,
model_directory=None,
result_directory=None,
real_device=False,
):
assert method in [
'qrac', 'qrac_zz', 'te', 'te_zz', 'zz_dis', 'zz_dis_cont'
], f"method {method} not exist"
if model_directory is None:
model_directory = f'models/Titanic_{method}_{epochs}_{positive_factor}_{depth}_{seed}_{reg}_model'
if result_directory is None:
result_directory = f'results/Titanic_{method}_{epochs}_{positive_factor}_{depth}_{seed}_{reg}_result'
all_discrete = method in ['qrac', 'te', 'zz_dis']
df_train, _, y_train, _ = load_titanic_pd('data/Titanic_train.csv',
'data/Titanic_test.csv',
as_category=all_discrete)
# Get mvp column
mvp_col = ['Sex', 'Age', 'Pclass', 'Fare']
df_train = df_train[mvp_col]
vqc_gen = None
feature_map = None
var_form = None
X_train = None
if method in ['zz_dis']:
X_train = df_train.values
feature_map = ZZFeatureMap(4)
var_form = RYRZ(4, depth=depth)
vqc_gen = VQC
if method in ['zz_dis_cont']:
df_train['Fare'] = np.log(df_train['Fare'] + 1)
df_train['Age'] = df_train['Age'] / 60
X_train = df_train.values
feature_map = ZZFeatureMap(4)
var_form = RYRZ(4, depth=depth)
vqc_gen = VQC
if method in ['qrac', 'te']:
X_train = binary_encoder(df_train.values)
print(X_train)
num_qubit = len(X_train[0]) // 3
if method == 'qrac':
feature_map = CustomFeatureMap('ALL3in1', 1, num_qubit)
var_form = RYRZ(num_qubit, depth=depth)
vqc_gen = VQC
if method == 'te':
feature_map = QuantumCircuit(num_qubit)
var_form = MyRYRZ(num_qubit, depth=depth)
vqc_gen = MyVQC
if method in ['qrac_zz']:
df_train['Fare'] = np.log(df_train['Fare'] + 1)
df_train['Age'] = df_train['Age'] / 60
X_train_num = U3gate_input_encoder(df_train[['Sex', 'Pclass']].values)
X_train_con = df_train[['Age', 'Fare']].values
X_train = np.concatenate([X_train_con, X_train_num], axis=1)
X = [Parameter(f'x[{i}]') for i in range(X_train_num.shape[1])]
qr = QuantumRegister(1, 'cat')
qc = QuantumCircuit(qr)
for i in range(1):
qc.u3(X[2 * i], X[2 * i + 1], 0, i)
feature_map_cat = qc
X1 = Parameter('x[2]')
X2 = Parameter('x[3]')
feature_map_con = ZZFeatureMap(2)
feature_map_con = feature_map_con.assign_parameters([X1, X2])
feature_map = feature_map_con.combine(feature_map_cat)
var_form = RYRZ(3, depth=depth)
vqc_gen = VQC
if method in ['te_zz']:
df_train['Fare'] = np.log(df_train['Fare'] + 1)
df_train['Age'] = df_train['Age'] / 60
X_train_num = binary_encoder(df_train[['Sex', 'Pclass']].values)
X_train_con = df_train[['Age', 'Fare']].values
print(X_train_num[:, None])
print(X_train_con)
X_train = np.array([[x, y] for x, y in zip(X_train_num, X_train_con)])
X1 = Parameter('x[0]')
X2 = Parameter('x[1]')
feature_map_con = ZZFeatureMap(2)
feature_map_con = feature_map_con.assign_parameters([X1, X2])
qr = QuantumRegister(1, 'cat')
feature_map_num = QuantumCircuit(qr)
feature_map = feature_map_num.combine(feature_map_con)
var_form = MyRYRZ_zz(1, 2, depth=depth)
vqc_gen = MyVQC_zz
assert feature_map is not None, "Feature map is none"
assert var_form is not None, "Varform is none"
if real_device:
# Please fix these line...
provider = IBMQ.get_provider()
backend = provider.get_backend('ibmq_london')
else:
backend = QasmSimulator({"method": "statevector_gpu"})
def optimizer_gen():
return SPSA(epochs)
result = kfold_vqc(feature_map,
var_form,
backend,
optimizer_gen,
seed,
X_train,
y_train,
model_directory=model_directory,
result_directory=result_directory,
k=4,
positivedata_duplicate_ratio=positive_factor,
vqc_gen=vqc_gen)
return_result = {
'train_acc': result['Training accuracies (mean)'][-1],
'train_f1': result['Training F1 scores (mean)'][-1],
'test_acc': result['Test accuracies (mean)'][-1],
'test_f1': result['Test F1 scores (mean)'][-1]
}
return return_result
from sklearn import preprocessing
iris = datasets.load_iris()
# load iris and normalise
x = preprocessing.normalize(iris.data)
x1_train = x[0:49, :] # class A
x2_train = x[50:99, :] # class B
training_input = {'A': x1_train, 'B': x2_train}
class_labels = ['A', 'B']
blocks = 1
sv = Statevector.from_label('0' * n)
# circuit = QuantumCircuit(n)
feature_map = ZZFeatureMap(n, reps=2, entanglement='full')
var_form = RealAmplitudes(n, reps=blocks, entanglement='full')
circuit = feature_map.combine(var_form)
def get_data_dict(params, x):
"""Get the parameters dict for the circuit"""
parameters = {}
for i, p in enumerate(feature_map.ordered_parameters):
parameters[p] = x[i]
for i, p in enumerate(var_form.ordered_parameters):
parameters[p] = params[i]
return parameters
def assign_label(bit_string, class_labels):
hamming_weight = sum([int(k) for k in list(bit_string)])
is_odd_parity = hamming_weight & 1
