量子神经网络比较

Question

import sys
import os
import pennylane as qml
from pennylane import numpy as np
import tensorflow as tf
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler

import matplotlib

matplotlib.use('Agg') # Use non-interactive backend
import matplotlib.pyplot as plt

from tensorflow.keras.layers import Dense
from importance_sampling.training import ConstantTimeImportanceTraining
from importance_sampling.samplers import ConstantTimeSampler
from importance_sampling.datasets import InMemoryDataset
from importance_sampling.reweighting import BiasedReweightingPolicy
from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau, ModelCheckpoint

import logging
from tensorflow.keras.utils import to_categorical # 导入独热编码工具

Ensure TensorFlow uses float32

tf.keras.backend.set_floatx('float32')

Redirect stdout to capture print outputs

class Tee(object):
def init(self, *files):
self.files = files
self.encoding = getattr(files[0], 'encoding', 'utf-8') # Add encoding attribute

text
def write(self, obj):
    for f in self.files:
        f.write(obj)
        f.flush()

def flush(self):
    for f in self.files:
        f.flush()

output_file = open('依然使用独热编码和softmax.log', 'w', encoding='utf-8')
original_stdout = sys.stdout
tee = Tee(sys.stdout, output_file)
sys.stdout = tee

Configure Logger to use Tee

logger = logging.getLogger('my_sampler')
logger.setLevel(logging.INFO)

stream_handler = logging.StreamHandler(tee)
stream_handler.setLevel(logging.INFO)

formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
stream_handler.setFormatter(formatter)

Remove default handlers to avoid duplicate logs

if not logger.handlers:
logger.addHandler(stream_handler)
else:
logger.handlers = []
logger.addHandler(stream_handler)

Set Qiskit log level to WARNING to suppress INFO logs

logging.getLogger('qiskit').setLevel(logging.WARNING)

def save_plot(fig, filename):
fig.savefig(filename)
plt.close(fig)

def plot_history(history):
# Plot training and validation accuracy
fig, ax = plt.subplots(figsize=(10, 5))
ax.plot(history['accuracy'], label='Training Accuracy')
if 'val_accuracy' in history:
ax.plot(history['val_accuracy'], label='Validation Accuracy')
else:
print("Validation accuracy not found in history.")
ax.set_title('Model Accuracy')
ax.set_ylabel('Accuracy')
ax.set_xlabel('Epoch')
ax.legend(loc='upper left')
save_plot(fig, 'accuracy_plot.png')

text
# Plot training and validation loss
fig, ax = plt.subplots(figsize=(10, 5))
ax.plot(history['loss'], label='Training Loss')
if 'val_loss' in history:
    ax.plot(history['val_loss'], label='Validation Loss')
else:
    print("Validation loss not found in history.")
ax.set_title('Model Loss')
ax.set_ylabel('Loss')
ax.set_xlabel('Epoch')
ax.legend(loc='upper right')
save_plot(fig, 'loss_plot.png')

def calculate_accuracy_fluctuation(history):
if 'val_accuracy' in history:
accuracies = history['val_accuracy']
elif 'accuracy' in history:
accuracies = history['accuracy']
else:
print("Accuracy metric not found in history.")
return None, None
mean_accuracy = np.mean(accuracies)
fluctuation = np.sum(np.abs(accuracies - mean_accuracy))
return mean_accuracy, fluctuation

Define a quantum device with 7 qubits

dev = qml.device("default.qubit", wires=7)

Define encoding functions

def encode_group(control_qubits, data_qubits, angle_q2_values, angle_q3_values):
qml.ctrl(qml.RY(angle_q2_values, wires=data_qubits['q2']), control=control_qubits, control_values=0)
qml.ctrl(qml.RY(angle_q2_values, wires=data_qubits['q2']), control=control_qubits, control_values=1)

text
qml.ctrl(qml.RY(angle_q3_values[0], wires=data_qubits['q3']), control=control_qubits, control_values=0)
qml.ctrl(qml.RY(angle_q3_values[1], wires=data_qubits['q3']), control=control_qubits, control_values=1)

def encode_group_second(control_qubits, data_qubits, angle_q3_values):
qml.ctrl(qml.RY(angle_q3_values[0], wires=data_qubits['q3']), control=control_qubits, control_values=0)
qml.ctrl(qml.RY(angle_q3_values[1], wires=data_qubits['q3']), control=control_qubits, control_values=1)

Definition of swap gate

def apply_controlled_swap(swap_qubit, q_swap1, q_swap2):
"""
Modular method to apply a controlled SWAP gate.

text
Parameters:
- swap_qubit (int): Control qubit for the SWAP gate.
- q_swap1 (int): First qubit to swap.
- q_swap2 (int): Second qubit to swap.
"""
qml.Hadamard(wires=swap_qubit)
qml.CSWAP(wires=[swap_qubit, q_swap1, q_swap2])
qml.Hadamard(wires=swap_qubit)

@qml.qnode(dev, interface='tf', diff_method='backprop')
def quantum_neural_network(inputs, encoding_weights, rot_weights, entangle_weights):
control_qubits = 0
qml.Hadamard(wires=0)
# Neuron groups
encoding_data = [
{'data_qubits': {'q2': 1, 'q3': 2}, 'swap_qubit': 3},
{'data_qubits': {'q2': 4, 'q3': 5}, 'swap_qubit': 6},
]

text
# Angle encoding: inputs are mapped to [0, pi/2]
angles_q2_list = inputs  # inputs shape is (2,) for binary classification

# Encoding weights mapped to angles_q3
angles_q3_list = encoding_weights  # shape: (2, n_layers, 2)

# 第一层：同时编码values和weights
for idx, data_set in enumerate(encoding_data):
    swap_qubit = data_set['swap_qubit']
    data_qubits = data_set['data_qubits']

    angle_q2 = angles_q2_list[idx]  # scalar value
    angle_q3_values = angles_q3_list[idx, 0, :]  # shape: (2,)
    # 编码一个神经元
    encode_group(control_qubits, data_qubits, angle_q2, angle_q3_values)

# 每层编码旋转门和纠缠门
for qubit in range(rot_weights.shape[1]):
    rot_angles = rot_weights[0, qubit, :]  # shape: (3,)
    qml.Rot(rot_angles[0], rot_angles[1], rot_angles[2], wires=qubit)
# Apply CRZ entanglement gates
for qubit in range(rot_weights.shape[1] - 1):
    qml.CRZ(entangle_weights[0, qubit], wires=[qubit, qubit + 1])

# 每层每个神经元应用一次swap操作
for data_set in encoding_data:
    swap_qubit = data_set['swap_qubit']
    data_qubits = data_set['data_qubits']
    apply_controlled_swap(swap_qubit, data_qubits['q3'], data_qubits['q2'])

# 第二层：仅编码weights
for layer in range(rot_weights.shape[0] - 1):  # 每层的内容
    layer_second = layer + 1
    # 每层完整编码特征值和权重值

    for idx, data_set in enumerate(encoding_data):
        data_qubits = data_set['data_qubits']

        angle_q3_values = angles_q3_list[idx, layer_second, :]  # shape: (2,)
        # 编码一个神经元
        encode_group_second(control_qubits, data_qubits, angle_q3_values)

    # 每层编码旋转门和纠缠门
    for qubit in range(rot_weights.shape[1]):
        rot_angles = rot_weights[layer_second, qubit, :]  # shape: (3,)
        qml.Rot(rot_angles[0], rot_angles[1], rot_angles[2], wires=qubit)
    # Apply CRZ entanglement gates
    for qubit in range(rot_weights.shape[1] - 1):
        qml.CRZ(entangle_weights[layer_second, qubit], wires=[qubit, qubit + 1])

    # 每层每个神经元应用一次swap操作
    for data_set in encoding_data:
        swap_qubit = data_set['swap_qubit']
        data_qubits = data_set['data_qubits']
        if layer % 2 == 0:
            apply_controlled_swap(data_qubits['q2'], swap_qubit, data_qubits['q3'])
        else:
            apply_controlled_swap(swap_qubit, data_qubits['q2'], data_qubits['q3'])

# Determine measured qubits based on number of layers
if rot_weights.shape[0] % 2 == 1:
    measured_qubits = [3, 6]
else:
    measured_qubits = [1, 4]

# Ensure that measured_qubits is not empty
if not measured_qubits:
    raise ValueError("No qubits selected for measurement.")

# Return expectation values
results = [qml.expval(qml.PauliZ(qubit)) for qubit in measured_qubits]
return results

Define the quantum neural network layer

class QuantumNeuralNetworkLayer(tf.keras.layers.Layer):
def init(self, n_qubits, n_layers, num_classes):
super(QuantumNeuralNetworkLayer, self).init()
self.n_qubits = n_qubits
self.n_layers = n_layers
self.num_classes = num_classes

text
    # Define encoding weights with L2 regularization
    initializer = tf.keras.initializers.GlorotUniform()
    regularizer = tf.keras.regularizers.l2(0.001)  # Reduced regularization

    self.encoding_weights = self.add_weight(
        shape=(2, self.n_layers, 2),
        initializer=initializer,
        regularizer=regularizer,
        trainable=True,
        name='encoding_weights',
        dtype=tf.float32
    )

    # Define rotation layer weights with L2 regularization
    self.rot_weights = self.add_weight(
        shape=(n_layers, n_qubits, 3),
        initializer=initializer,
        regularizer=regularizer,
        trainable=True,
        name='rot_weights',
        dtype=tf.float32
    )

    # Define entangling weights with L2 regularization
    self.entangle_weights = self.add_weight(
        shape=(n_layers, n_qubits - 1),
        initializer=tf.keras.initializers.RandomUniform(0, 2 * np.pi),
        regularizer=regularizer,
        trainable=True,
        name='entangle_weights',
        dtype=tf.float32
    )

def call(self, inputs):
    outputs_list = []
    for i in range(tf.shape(inputs)[0]):
        x = inputs[i]
        x = tf.cast(x, tf.float32)
        result = quantum_neural_network(x, self.encoding_weights, self.rot_weights, self.entangle_weights)
        outputs_list.append(tf.cast(result, tf.float32))

    outputs = tf.stack(outputs_list, axis=0)
    return outputs

Define the quantum neural network model with enhanced classical layers

class QuantumNeuralNetwork(tf.keras.Model):
def init(self, n_qubits, n_layers, num_classes, **kwargs):
super(QuantumNeuralNetwork, self).init(**kwargs)
self.n_qubits = n_qubits
self.n_layers = n_layers
self.num_classes = num_classes

text
    # 量子层
    self.quantum_layer = QuantumNeuralNetworkLayer(n_qubits, n_layers, num_classes)

    # 经典层，使用层归一化代替批归一化
    self.dense1 = tf.keras.layers.Dense(
        64,
        activation='relu',
        kernel_regularizer=tf.keras.regularizers.l2(0.00001),
        dtype=tf.float32
    )
    self.layer_norm1 = tf.keras.layers.BatchNormalization()

    self.dense2 = tf.keras.layers.Dense(
        32,
        activation='relu',
        kernel_regularizer=tf.keras.regularizers.l2(0.00001),
        dtype=tf.float32
    )
    self.layer_norm2 = tf.keras.layers.BatchNormalization()

    # 如果需要，可以取消注释以下层并相应地添加层归一化
    # self.dense3 = tf.keras.layers.Dense(
    #     16,
    #     activation='relu',
    #     kernel_regularizer=tf.keras.regularizers.l2(0.001),
    #     dtype=tf.float32
    # )
    # self.layer_norm3 = tf.keras.layers.LayerNormalization()

    # 输出层，使用 softmax 激活
    self.output_layer = tf.keras.layers.Dense(
        num_classes,
        activation='softmax',
        kernel_regularizer=tf.keras.regularizers.l2(0.0001),
        dtype=tf.float32
    )

def call(self, inputs, training=False):
    # 通过量子层
    q_output = self.quantum_layer(inputs)

    # 第一层全连接 + 层归一化
    x = self.dense1(q_output)
    x = self.layer_norm1(x, training=training)

    # 第二层全连接 + 层归一化
    x = self.dense2(x)
    x = self.layer_norm2(x, training=training)

    # 如果使用第三层，可以取消注释以下代码
    # x = self.dense3(x)
    # x = self.layer_norm3(x, training=training)

    # 输出层
    output = self.output_layer(x)
    return output

def is_quantum_model(model):
for layer in model.layers:
if isinstance(layer, QuantumNeuralNetworkLayer):
return True
elif hasattr(layer, 'layers'):
if is_quantum_model(layer):
return True
return False

if name == "main":
try:
# Load data
data = np.load('plane_point_set.npz')
X = data['points'].astype(np.float32) # Original coordinates
phase_angles = data['phase_angles'].astype(np.float32) # Phase angles mapped to [0, π/2]
y = data['labels'].reshape(-1, 1)

text
    # 将标签转换为独热编码
    y_encoded = to_categorical(y, num_classes=2).astype(np.float32)  # 修改为独热编码
    num_classes = 2  # 修改为2类

    print("num_classes", num_classes)

    # Split dataset into training and testing
    x_full_train, x_test, phase_full_train, phase_test, y_full_train, y_test = train_test_split(
        X, phase_angles, y_encoded, test_size=0.2, random_state=116, stratify=y_encoded
    )

    # Further split training set into training and validation sets
    x_train, x_val, phase_train, phase_val, y_train, y_val = train_test_split(
        x_full_train, phase_full_train, y_full_train,
        test_size=0.25, random_state=42, stratify=y_full_train
    )

    # Shuffle the training data manually
    permutation = np.random.permutation(len(x_train))
    x_train = x_train[permutation]
    phase_train = phase_train[permutation]
    y_train = y_train[permutation]

    # Hyperparameters
    learning_rate = 0.0005  # Reduced learning rate
    epochs = 3  # Increased number of epochs
    batch_size = 128  # Adjusted batch size
    n_qubits = 7
    n_layers = 2
    num_classes = 2  # Binary classification with one-hot encoding

    model = QuantumNeuralNetwork(
        n_qubits=n_qubits,
        n_layers=n_layers,
        num_classes=num_classes
    )

    if is_quantum_model(model):
        print("The model contains quantum neural network (QNN) layers.")
    else:
        print("The model does not contain quantum neural network (QNN) layers.")

    # Draw the quantum circuit
    sample_inputs = np.ones(2, dtype=np.float32) * (np.pi / 4)  # Inputs mapped to [0, π/2]
    sample_encoding_weights = np.zeros((2, n_layers, 2), dtype=np.float32)
    sample_rot_weights = np.zeros((n_layers, n_qubits, 3), dtype=np.float32)
    sample_entangle_weights = np.zeros((n_layers, n_qubits - 1), dtype=np.float32)

    qml.drawer.use_style('sketch')
    print("\nQuantum circuit diagram:")
    try:
        circuit_text = qml.draw(quantum_neural_network)(sample_inputs, sample_encoding_weights, sample_rot_weights,
                                                        sample_entangle_weights)
        print(circuit_text)
    except Exception as e:
        print(f"Error drawing the quantum circuit: {e}")

    try:
        fig, ax = qml.draw_mpl(quantum_neural_network)(sample_inputs, sample_encoding_weights, sample_rot_weights,
                                                       sample_entangle_weights)
        fig.savefig('quantum_circuit.png')
    except Exception as e:
        print(f"Error saving the quantum circuit diagram: {e}")

    steps_per_epoch = len(x_train) // batch_size
    total_steps = epochs * steps_per_epoch

    # Create dataset (including validation set)
    try:
        dataset = InMemoryDataset(
            x_train,
            y_train,
            x_val,
            y_val,
            categorical=True  # 修改为True以支持独热编码
            # Removed shuffle=True as it's not supported
        )
    except TypeError as e:
        print(f"Error initializing InMemoryDataset: {e}")
        print("Attempting to initialize without 'shuffle' parameter.")
        dataset = InMemoryDataset(
            x_train,
            y_train,
            x_val,
            y_val,
            categorical=True  # 保持为True
        )
    except Exception as e:
        print(f"Unexpected error initializing InMemoryDataset: {e}")
        sys.exit(1)

    # Initialize reweighting policy
    reweighting_policy = BiasedReweightingPolicy(k=1.0)

    # Initialize importance sampling training
    wrapped_model = ConstantTimeImportanceTraining(model, score="loss")  # Use loss as score

    # Modify sampler to use dynamic alpha and beta
    def custom_sampler(dataset, batch_size, steps_per_epoch, epochs):
        return ConstantTimeSampler(
            dataset,
            reweighting_policy,
            model=wrapped_model.model,
        )

    # Assign custom sampler to wrapped_model
    wrapped_model.sampler = custom_sampler

    # Use CategoricalCrossentropy loss function
    loss_fn = tf.keras.losses.CategoricalCrossentropy()

    # Compile model
    model.compile(
        optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate),
        loss=loss_fn,
        metrics=['accuracy'],
        run_eagerly=True  # Set to False for better performance after debugging
    )


    # Create a callback class to save training metrics
    class MetricsHistory(tf.keras.callbacks.Callback):
        def on_train_begin(self, logs=None):
            self.history = {'loss': [], 'accuracy': [], 'val_loss': [], 'val_accuracy': []}

        def on_epoch_end(self, epoch, logs=None):
            self.history['loss'].append(logs.get('loss'))
            self.history['accuracy'].append(logs.get('accuracy'))
            self.history['val_loss'].append(logs.get('val_loss'))
            self.history['val_accuracy'].append(logs.get('val_accuracy'))


    # Instantiate callbacks
    metrics_history = MetricsHistory()

    # Add callbacks
    callbacks = [
        metrics_history,
        EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True),
        ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=5, min_lr=1e-5),
        ModelCheckpoint('best_model.h5', save_best_only=True, monitor='val_loss')
    ]

    # Train model using fit_dataset
    try:
        history = wrapped_model.fit_dataset(
            dataset=dataset,
            batch_size=batch_size,
            epochs=epochs,
            steps_per_epoch=steps_per_epoch,
            verbose=1,
            callbacks=callbacks
        )
    except Exception as e:
        print(f"Error during training: {e}")

    # Evaluate model on test set
        # 将测试集标签转换为独热编码格式
    y_test_encoded = to_categorical(y_test, num_classes=2).astype(np.float32)
    print("\nEvaluating on test set:")
    test_loss, test_accuracy = wrapped_model.evaluate((x_test, y_test_encoded), mode='test')
    print('Test loss:', test_loss)
    print('Test accuracy:', test_accuracy)
    # Plot graphs using history from custom callback
    plot_history(metrics_history.history)

    mean_accuracy, fluctuation = calculate_accuracy_fluctuation(metrics_history.history)
    if mean_accuracy is not None and fluctuation is not None:
        print('Validation set mean accuracy:', mean_accuracy)
        print('Accuracy fluctuation:', fluctuation)

except Exception as e:
    print(f"An unexpected error occurred: {e}")
finally:
    # Close output file and restore stdout
    sys.stdout = original_stdout
    output_file.close()

print(
    "Script execution completed. Please check '不重复编码Ry-翻转swap.log' for print output and check the saved plots in the current directory.")

上述是代码1
import sys
import os
import pennylane as qml
from pennylane import numpy as np
import tensorflow as tf
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler

import matplotlib

matplotlib.use('Agg') # Use non-interactive backend
import matplotlib.pyplot as plt

from tensorflow.keras.layers import Dense
from importance_sampling.training import ConstantTimeImportanceTraining
from importance_sampling.samplers import ConstantTimeSampler
from importance_sampling.datasets import InMemoryDataset
from importance_sampling.reweighting import BiasedReweightingPolicy
from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau, ModelCheckpoint

import logging

Ensure TensorFlow uses float32

tf.keras.backend.set_floatx('float32')

Redirect stdout to capture print outputs

class Tee(object):
def init(self, *files):
self.files = files
self.encoding = getattr(files[0], 'encoding', 'utf-8') # Add encoding attribute

text
def write(self, obj):
    for f in self.files:
        f.write(obj)
        f.flush()

def flush(self):
    for f in self.files:
        f.flush()

output_file = open('缩减量子层.log', 'w', encoding='utf-8')
original_stdout = sys.stdout
tee = Tee(sys.stdout, output_file)
sys.stdout = tee

Configure Logger to use Tee

logger = logging.getLogger('my_sampler')
logger.setLevel(logging.INFO)

stream_handler = logging.StreamHandler(tee)
stream_handler.setLevel(logging.INFO)

formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
stream_handler.setFormatter(formatter)

Remove default handlers to avoid duplicate logs

if not logger.handlers:
logger.addHandler(stream_handler)
else:
logger.handlers = []
logger.addHandler(stream_handler)

Set Qiskit log level to WARNING to suppress INFO logs

logging.getLogger('qiskit').setLevel(logging.WARNING)

def save_plot(fig, filename):
fig.savefig(filename)
plt.close(fig)

def plot_history(history):
# Plot training and validation accuracy
fig, ax = plt.subplots(figsize=(10, 5))
ax.plot(history['accuracy'], label='Training Accuracy')
if 'val_accuracy' in history:
ax.plot(history['val_accuracy'], label='Validation Accuracy')
else:
print("Validation accuracy not found in history.")
ax.set_title('Model Accuracy')
ax.set_ylabel('Accuracy')
ax.set_xlabel('Epoch')
ax.legend(loc='upper left')
save_plot(fig, 'accuracy_plot.png')

text
# Plot training and validation loss
fig, ax = plt.subplots(figsize=(10, 5))
ax.plot(history['loss'], label='Training Loss')
if 'val_loss' in history:
    ax.plot(history['val_loss'], label='Validation Loss')
else:
    print("Validation loss not found in history.")
ax.set_title('Model Loss')
ax.set_ylabel('Loss')
ax.set_xlabel('Epoch')
ax.legend(loc='upper right')
save_plot(fig, 'loss_plot.png')

def calculate_accuracy_fluctuation(history):
if 'val_accuracy' in history:
accuracies = history['val_accuracy']
elif 'accuracy' in history:
accuracies = history['accuracy']
else:
print("Accuracy metric not found in history.")
return None, None
mean_accuracy = np.mean(accuracies)
fluctuation = np.sum(np.abs(accuracies - mean_accuracy))
return mean_accuracy, fluctuation

Define a quantum device with 7 qubits

dev = qml.device("default.qubit", wires=7)

Define encoding functions

def encode_group(control_qubits, data_qubits, angle_q2_values, angle_q3_values):
qml.ctrl(qml.RY(angle_q2_values, wires=data_qubits['q2']), control=control_qubits, control_values=0)
qml.ctrl(qml.RY(angle_q2_values, wires=data_qubits['q2']), control=control_qubits, control_values=1)

text
qml.ctrl(qml.RY(angle_q3_values[0], wires=data_qubits['q3']), control=control_qubits, control_values=0)
qml.ctrl(qml.RY(angle_q3_values[1], wires=data_qubits['q3']), control=control_qubits, control_values=1)

def encode_group_second(control_qubits, data_qubits, angle_q3_values):
qml.ctrl(qml.RY(angle_q3_values[0], wires=data_qubits['q3']), control=control_qubits, control_values=0)
qml.ctrl(qml.RY(angle_q3_values[1], wires=data_qubits['q3']), control=control_qubits, control_values=1)

Definition of swap gate

def apply_controlled_swap(swap_qubit, q_swap1, q_swap2):
"""
Modular method to apply a controlled SWAP gate.

text
Parameters:
- swap_qubit (int): Control qubit for the SWAP gate.
- q_swap1 (int): First qubit to swap.
- q_swap2 (int): Second qubit to swap.
"""
qml.Hadamard(wires=swap_qubit)
qml.CSWAP(wires=[swap_qubit, q_swap1, q_swap2])
qml.Hadamard(wires=swap_qubit)

@qml.qnode(dev, interface='tf', diff_method='backprop')
def quantum_neural_network(inputs, encoding_weights, rot_weights, entangle_weights):
control_qubits = 0
qml.Hadamard(wires=0)
# Neuron groups
encoding_data = [
{'data_qubits': {'q2': 1, 'q3': 2}, 'swap_qubit': 3},
{'data_qubits': {'q2': 4, 'q3': 5}, 'swap_qubit': 6},
]

text
# Angle encoding: inputs are mapped to [0, pi/2]
angles_q2_list = inputs  # inputs shape is (2,) for binary classification

# Encoding weights mapped to angles_q3
angles_q3_list = encoding_weights  # shape: (2, n_layers, 2)

# 第一层：同时编码values和weights
for idx, data_set in enumerate(encoding_data):
    swap_qubit = data_set['swap_qubit']
    data_qubits = data_set['data_qubits']

    angle_q2 = angles_q2_list[idx]  # scalar value
    angle_q3_values = angles_q3_list[idx, 0, :]  # shape: (2,)
    # 编码一个神经元
    encode_group(control_qubits, data_qubits, angle_q2, angle_q3_values)

# 每层编码旋转门和纠缠门
for qubit in range(rot_weights.shape[1]):
    rot_angles = rot_weights[0, qubit, :]  # shape: (3,)
    qml.Rot(rot_angles[0], rot_angles[1], rot_angles[2], wires=qubit)
# # Apply CRZ entanglement gates
# for qubit in range(rot_weights.shape[1] - 1):
#     qml.CRZ(entangle_weights[0, qubit], wires=[qubit, qubit + 1])

# 每层每个神经元应用一次swap操作
for data_set in encoding_data:
    swap_qubit = data_set['swap_qubit']
    data_qubits = data_set['data_qubits']
    apply_controlled_swap(swap_qubit, data_qubits['q3'], data_qubits['q2'])

# 第二层：仅编码weights
for layer in range(rot_weights.shape[0] - 1):  # 每层的内容
    layer_second = layer + 1
    # 每层完整编码特征值和权重值

    for idx, data_set in enumerate(encoding_data):
        data_qubits = data_set['data_qubits']

        angle_q3_values = angles_q3_list[idx, layer_second, :]  # shape: (2,)
        # 编码一个神经元
        encode_group_second(control_qubits, data_qubits, angle_q3_values)

    # 每层编码旋转门和纠缠门
    for qubit in range(rot_weights.shape[1]):
        rot_angles = rot_weights[layer_second, qubit, :]  # shape: (3,)
        qml.Rot(rot_angles[0], rot_angles[1], rot_angles[2], wires=qubit)
    # # Apply CRZ entanglement gates
    # for qubit in range(rot_weights.shape[1] - 1):
    #     qml.CRZ(entangle_weights[layer_second, qubit], wires=[qubit, qubit + 1])

    # 每层每个神经元应用一次swap操作
    for data_set in encoding_data:
        swap_qubit = data_set['swap_qubit']
        data_qubits = data_set['data_qubits']
        if layer % 2 == 0:
            apply_controlled_swap(data_qubits['q2'], swap_qubit, data_qubits['q3'])
        else:
            apply_controlled_swap(swap_qubit, data_qubits['q2'], data_qubits['q3'])

# Determine measured qubits based on number of layers
if rot_weights.shape[0] % 2 == 1:
    measured_qubits = [3, 6]
else:
    measured_qubits = [1, 4]

# Ensure that measured_qubits is not empty
if not measured_qubits:
    raise ValueError("No qubits selected for measurement.")

# Return expectation values
results = [qml.expval(qml.PauliZ(qubit)) for qubit in measured_qubits]
return results

Define the quantum neural network layer

class QuantumNeuralNetworkLayer(tf.keras.layers.Layer):
def init(self, n_qubits, n_layers, num_classes):
super(QuantumNeuralNetworkLayer, self).init()
self.n_qubits = n_qubits
self.n_layers = n_layers
self.num_classes = num_classes

text
    # Define encoding weights with L2 regularization
    initializer = tf.keras.initializers.GlorotUniform()
    regularizer = tf.keras.regularizers.l2(0.001)  # Reduced regularization

    self.encoding_weights = self.add_weight(
        shape=(2, self.n_layers, 2),
        initializer=initializer,
        regularizer=regularizer,
        trainable=True,
        name='encoding_weights',
        dtype=tf.float32
    )

    # Define rotation layer weights with L2 regularization
    self.rot_weights = self.add_weight(
        shape=(n_layers, n_qubits, 3),
        initializer=initializer,
        regularizer=regularizer,
        trainable=True,
        name='rot_weights',
        dtype=tf.float32
    )

    # Define entangling weights with L2 regularization
    self.entangle_weights = self.add_weight(
        shape=(n_layers, n_qubits - 1),
        initializer=tf.keras.initializers.RandomUniform(0, 2 * np.pi),
        regularizer=regularizer,
        trainable=True,
        name='entangle_weights',
        dtype=tf.float32
    )

def call(self, inputs):
    outputs_list = []
    for i in range(tf.shape(inputs)[0]):
        x = inputs[i]
        x = tf.cast(x, tf.float32)
        result = quantum_neural_network(x, self.encoding_weights, self.rot_weights, self.entangle_weights)
        outputs_list.append(tf.cast(result, tf.float32))

    outputs = tf.stack(outputs_list, axis=0)
    return outputs

Define the quantum neural network model with enhanced classical layers

class QuantumNeuralNetwork(tf.keras.Model):
def init(self, n_qubits, n_layers, num_classes, **kwargs):
super(QuantumNeuralNetwork, self).init(**kwargs)
self.n_qubits = n_qubits
self.n_layers = n_layers
self.num_classes = num_classes

text
    # Quantum layer
    self.quantum_layer = QuantumNeuralNetworkLayer(n_qubits, n_layers, num_classes)

    # Classical layers with increased capacity and batch normalization
    self.dense1 = tf.keras.layers.Dense(
        64,
        activation='relu',
        kernel_regularizer=tf.keras.regularizers.l2(0.00001),
        dtype=tf.float32
    )
    # self.batch_norm1 = tf.keras.layers.BatchNormalization()

    self.dense2 = tf.keras.layers.Dense(
        32,
        activation='relu',
        kernel_regularizer=tf.keras.regularizers.l2(0.00001),
        dtype=tf.float32
    )
    # self.batch_norm2 = tf.keras.layers.BatchNormalization()

    # self.dense3 = tf.keras.layers.Dense(
    #     16,
    #     activation='relu',
    #     kernel_regularizer=tf.keras.regularizers.l2(0.001),
    #     dtype=tf.float32
    # )
    # self.batch_norm3 = tf.keras.layers.BatchNormalization()

    # Output layer for binary classification
    self.output_layer = tf.keras.layers.Dense(
        1,
        activation='sigmoid',
        kernel_regularizer=tf.keras.regularizers.l2(0.001),
        dtype=tf.float32
    )

def call(self, inputs):
    q_output = self.quantum_layer(inputs)
    x = self.dense1(q_output)
    # x = self.batch_norm1(x)
    x = self.dense2(x)
    # x = self.batch_norm2(x)
    # x = self.dense3(x)
    # x = self.batch_norm3(x)
    return self.output_layer(x)

def is_quantum_model(model):
for layer in model.layers:
if isinstance(layer, QuantumNeuralNetworkLayer):
return True
elif hasattr(layer, 'layers'):
if is_quantum_model(layer):
return True
return False

if name == "main":
try:
# Load data
data = np.load('plane_point_set.npz')
X = data['points'].astype(np.float32) # Original coordinates
phase_angles = data['phase_angles'].astype(np.float32) # Phase angles mapped to [0, π/2]
y = data['labels'].reshape(-1, 1)

text
    # For BinaryCrossentropy, labels should be single binary values
    y_encoded = y.astype(np.float32)  # No need for OneHotEncoder
    num_classes = 1  # Binary classification
    print("num_classes", num_classes)

    # Split dataset into training and testing
    x_full_train, x_test, phase_full_train, phase_test, y_full_train, y_test = train_test_split(
        X, phase_angles, y_encoded, test_size=0.2, random_state=116, stratify=y_encoded
    )

    # Further split training set into training and validation sets
    x_train, x_val, phase_train, phase_val, y_train, y_val = train_test_split(
        x_full_train, phase_full_train, y_full_train,
        test_size=0.25, random_state=42, stratify=y_full_train
    )

    # Shuffle the training data manually
    permutation = np.random.permutation(len(x_train))
    x_train = x_train[permutation]
    phase_train = phase_train[permutation]
    y_train = y_train[permutation]

    # Hyperparameters
    learning_rate = 0.001  # Reduced learning rate
    epochs = 5  # Increased number of epochs
    batch_size = 128  # Adjusted batch size
    n_qubits = 7
    n_layers = 3
    num_classes = 1  # Binary classification

    model = QuantumNeuralNetwork(
        n_qubits=n_qubits,
        n_layers=n_layers,
        num_classes=num_classes
    )

    if is_quantum_model(model):
        print("The model contains quantum neural network (QNN) layers.")
    else:
        print("The model does not contain quantum neural network (QNN) layers.")

    # Draw the quantum circuit
    sample_inputs = np.ones(2, dtype=np.float32) * (np.pi / 4)  # Inputs mapped to [0, π/2]
    sample_encoding_weights = np.zeros((2, n_layers, 2), dtype=np.float32)
    sample_rot_weights = np.zeros((n_layers, n_qubits, 3), dtype=np.float32)
    sample_entangle_weights = np.zeros((n_layers, n_qubits - 1), dtype=np.float32)

    qml.drawer.use_style('sketch')
    print("\nQuantum circuit diagram:")
    try:
        circuit_text = qml.draw(quantum_neural_network)(sample_inputs, sample_encoding_weights, sample_rot_weights,
                                                        sample_entangle_weights)
        print(circuit_text)
    except Exception as e:
        print(f"Error drawing the quantum circuit: {e}")

    try:
        fig, ax = qml.draw_mpl(quantum_neural_network)(sample_inputs, sample_encoding_weights, sample_rot_weights,
                                                       sample_entangle_weights)
        fig.savefig('quantum_circuit.png')
    except Exception as e:
        print(f"Error saving the quantum circuit diagram: {e}")

    steps_per_epoch = len(x_train) // batch_size
    total_steps = epochs * steps_per_epoch

    # Create dataset (including validation set)
    try:
        dataset = InMemoryDataset(
            x_train,
            y_train,
            x_val,
            y_val,
            categorical=False  # Changed to False for BinaryCrossentropy
            # Removed shuffle=True as it's not supported
        )
    except TypeError as e:
        print(f"Error initializing InMemoryDataset: {e}")
        print("Attempting to initialize without 'shuffle' parameter.")
        dataset = InMemoryDataset(
            x_train,
            y_train,
            x_val,
            y_val,
            categorical=False
        )
    except Exception as e:
        print(f"Unexpected error initializing InMemoryDataset: {e}")
        sys.exit(1)

    # Initialize reweighting policy
    reweighting_policy = BiasedReweightingPolicy(k=1.0)

    # Initialize importance sampling training
    wrapped_model = ConstantTimeImportanceTraining(model, score="loss")  # Use loss as score


    # Modify sampler to use dynamic alpha and beta
    def custom_sampler(dataset, batch_size, steps_per_epoch, epochs):
        return ConstantTimeSampler(
            dataset,
            reweighting_policy,
            model=wrapped_model.model,
        )


    # Assign custom sampler to wrapped_model
    wrapped_model.sampler = custom_sampler

    # Use BinaryCrossentropy loss function
    loss_fn = tf.keras.losses.BinaryCrossentropy()

    # Compile model
    model.compile(
        optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate),
        loss=loss_fn,
        metrics=['accuracy'],
        run_eagerly=True  # Set to False for better performance after debugging
    )


    # Create a callback class to save training metrics
    class MetricsHistory(tf.keras.callbacks.Callback):
        def on_train_begin(self, logs=None):
            self.history = {'loss': [], 'accuracy': [], 'val_loss': [], 'val_accuracy': []}

        def on_epoch_end(self, epoch, logs=None):
            self.history['loss'].append(logs.get('loss'))
            self.history['accuracy'].append(logs.get('accuracy'))
            self.history['val_loss'].append(logs.get('val_loss'))
            self.history['val_accuracy'].append(logs.get('val_accuracy'))


    # Instantiate callbacks
    metrics_history = MetricsHistory()

    # Add callbacks
    callbacks = [
        metrics_history,
        EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True),
        ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=5, min_lr=1e-5),
        ModelCheckpoint('best_model.h5', save_best_only=True, monitor='val_loss')
    ]

    # Train model using fit_dataset
    try:
        history = wrapped_model.fit_dataset(
            dataset=dataset,
            batch_size=batch_size,
            epochs=epochs,
            steps_per_epoch=steps_per_epoch,
            verbose=1,
            callbacks=callbacks
        )
    except Exception as e:
        print(f"Error during training: {e}")

    # Evaluate model on test set
    try:
        score = wrapped_model.evaluate((x_test, y_test))
        print('Test loss:', score[0])
        print('Test accuracy:', score[1])
    except Exception as e:
        print(f"Error during evaluation: {e}")

    # Plot graphs using history from custom callback
    plot_history(metrics_history.history)

    mean_accuracy, fluctuation = calculate_accuracy_fluctuation(metrics_history.history)
    if mean_accuracy is not None and fluctuation is not None:
        print('Validation set mean accuracy:', mean_accuracy)
        print('Accuracy fluctuation:', fluctuation)

except Exception as e:
    print(f"An unexpected error occurred: {e}")
finally:
    # Close output file and restore stdout
    sys.stdout = original_stdout
    output_file.close()

print(
    "Script execution completed. Please check '不重复编码Ry-翻转swap.log' for print output and check the saved plots in the current directory.")

上述是代码2
告诉我代码1和代码2之间在神经网络之间的区别

Answer 1

下面从「量子层」和「经典层（即神经网络部分）」两方面，总结代码1与代码2在神经网络结构与用法上的主要差异。

1. 输出层与损失函数不同

代码1（多输出 + Softmax + CategoricalCrossentropy）
- 最后一层 self.output_layer = tf.keras.layers.Dense(num_classes, activation='softmax', ...)，num_classes = 2
- 配合使用 loss_fn = tf.keras.losses.CategoricalCrossentropy()
- 标签 y 采用了独热编码 (one-hot encoding)，即 $1,0$ 或 $0,1$ 形式
- 适合多分类（即使是二分类，也走的是“多分类”范式）
代码2（单输出 + Sigmoid + BinaryCrossentropy）
- 最后一层 self.output_layer = tf.keras.layers.Dense(1, activation='sigmoid', ...)，输出 0~1 之间的概率
- 配合使用 loss_fn = tf.keras.losses.BinaryCrossentropy()
- 标签 y 为 0 或 1 的浮点数（不再是独热编码）
- 典型的二分类范式

因此，代码1 的分类模式是“softmax+独热编码+交叉熵（多分类形式）”，而 代码2 则是“sigmoid+单输出+交叉熵（二分类形式）”。

2. 量子电路（Quantum Layer）规模与门操作的差异

代码1
- 保留了 qml.CRZ(entangle_weights[0, qubit], wires=[qubit, qubit + 1]) 等纠缠操作（在第一层和后续层都调用了 CRZ）
- 每一层都包含旋转门 qml.Rot(...) + CRZ 纠缠门，并且在循环中执行相应的 controlled-SWAP
代码2
- 部分纠缠门（CRZ）被注释掉了，等同于去掉了显式的 CRZ 纠缠操作，仅保留了各个 qml.Rot(...)
- 量子层数（n_layers）设置上也略有变化（代码里设置 n_layers = 3，但真正的纠缠操作被注释掉，实际变得更“轻量”或“缩减”）
- 仍然保留了对部分 qubit 的旋转编码，以及 controlled-SWAP 操作，但整体纠缠强度相比代码1明显降低

所以，代码2 的量子网络更“简化”或“缩减”，在量子态的纠缠程度上不如代码1那么复杂。

3. 经典网络结构及归一化层的差异

代码1（经典部分）
- 经典层：Dense(64, relu) -> BatchNormalization -> Dense(32, relu) -> BatchNormalization -> ... -> Dense(2, softmax)
- 即有两层全连接（每层后都有 Layer Normalization 或 BatchNormalization），最后再接输出层。
- 输出层神经元数=2，激活函数=softmax。
代码2（经典部分）
- 经典层：Dense(64, relu) -> Dense(32, relu) -> ... -> Dense(1, sigmoid)
- 将 BatchNormalization 相关代码大部分注释或简化掉了
- 输出层神经元数=1，激活函数=sigmoid

因此，代码2 的经典网络层数、正则化方式（BatchNormalization 等）与 代码1 不同，而且输出层结构也简化为单神经元输出。

4. 标签处理方式

代码1：y 转换为 one-hot，使用 to_categorical(..., num_classes=2)，并且 num_classes=2
代码2：y 直接保留为 0 或 1 的 float 值，不做 one-hot

5. 小结

输出层：代码1 是二分类但使用了“多类方式（2维输出 + softmax + categorical crossentropy）”；代码2 是单输出 + sigmoid + binary crossentropy 的典型二分类方式。
量子层：代码2 里注释掉了 CRZ 纠缠门，因此相比代码1 的量子电路，纠缠结构更弱，实际层数虽多但可表达能力也相对简化。
经典层：代码1 使用了带 BatchNormalization 的多层网络、最后是 2 维 softmax 输出；代码2 的经典层简化为 Dense(64) -> Dense(32) -> Dense(1) 且部分正则化层注释。
标签：代码1 使用独热编码、代码2 使用二值标签。

这几处即是代码1 与代码2 在神经网络（包括量子部分与经典部分）最主要的区别所在。

量子神经网络比较

Pregunta

Ensure TensorFlow uses float32

Redirect stdout to capture print outputs

Configure Logger to use Tee

Remove default handlers to avoid duplicate logs

Set Qiskit log level to WARNING to suppress INFO logs

Define a quantum device with 7 qubits

Define encoding functions

Definition of swap gate

Define the quantum neural network layer

Define the quantum neural network model with enhanced classical layers

Ensure TensorFlow uses float32

Redirect stdout to capture print outputs

Configure Logger to use Tee

Remove default handlers to avoid duplicate logs

Set Qiskit log level to WARNING to suppress INFO logs

Define a quantum device with 7 qubits

Define encoding functions

Definition of swap gate

Define the quantum neural network layer

Define the quantum neural network model with enhanced classical layers

Pensamiento

Respuesta

1. 输出层与损失函数不同

2. 量子电路（Quantum Layer）规模与门操作的差异

3. 经典网络结构及归一化层的差异

4. 标签处理方式

5. 小结

Compartir este Q&A