318 lines
12 KiB
Python
318 lines
12 KiB
Python
import torch.nn.functional as F
|
|
import numpy as np
|
|
import torch
|
|
import torch.nn as nn
|
|
from torch.utils.data import Dataset
|
|
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, confusion_matrix
|
|
import torch.optim as optim
|
|
from sklearn.preprocessing import StandardScaler
|
|
from torch.utils.tensorboard import SummaryWriter
|
|
from torch.cuda.amp import GradScaler, autocast
|
|
import os
|
|
from sklearn.metrics import precision_score, recall_score, f1_score
|
|
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
writer = SummaryWriter() # 初始化 TensorBoard
|
|
os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0'
|
|
# 自定义数据集，包含数据增强（添加噪声）
|
|
from skopt import BayesSearchCV
|
|
from skopt.space import Real, Integer
|
|
import torch
|
|
import torch.optim as optim
|
|
from torch.utils.data import DataLoader
|
|
class MyDataset(Dataset):
|
|
def __init__(self, specs, labels, augment=False):
|
|
self.specs = specs
|
|
self.labels = labels
|
|
self.augment = augment # 是否启用数据增强
|
|
|
|
def __getitem__(self, index):
|
|
spec, target = self.specs[index], self.labels[index]
|
|
|
|
# 数据增强：在训练数据上添加随机噪声
|
|
if self.augment:
|
|
noise = 0.01 * torch.randn_like(spec)
|
|
spec = spec + noise
|
|
|
|
return spec, target
|
|
|
|
def __len__(self):
|
|
return len(self.specs)
|
|
|
|
|
|
# 标准化数据
|
|
def ZspPocess(X_train, X_test, y_train, y_test, need=True):
|
|
if need:
|
|
scaler = StandardScaler()
|
|
X_train = scaler.fit_transform(X_train) # fit_transform 用于训练集
|
|
X_test = scaler.transform(X_test) # 只对测试集应用 transform
|
|
|
|
# 将标准化的数据转换为 Tensor
|
|
X_train = torch.tensor(X_train[:, np.newaxis, :], dtype=torch.float32)
|
|
X_test = torch.tensor(X_test[:, np.newaxis, :], dtype=torch.float32)
|
|
y_train = torch.tensor(y_train, dtype=torch.long)
|
|
y_test = torch.tensor(y_test, dtype=torch.long)
|
|
|
|
# 使用数据增强 (augment=True) 创建训练集
|
|
data_train = MyDataset(X_train, y_train, augment=True)
|
|
data_test = MyDataset(X_test, y_test, augment=False)
|
|
|
|
return data_train, data_test
|
|
|
|
|
|
# CNN 模型，添加 Dropout 层和调整 Dropout 率
|
|
class CNN3Layers(nn.Module):
|
|
def __init__(self, nls, dropout_conv=0.3, dropout_fc=0.5):
|
|
super(CNN3Layers, self).__init__()
|
|
self.CONV1 = nn.Sequential(
|
|
nn.Conv1d(1, 64, 5, 1, padding=2),
|
|
nn.BatchNorm1d(64),
|
|
nn.ReLU(),
|
|
nn.MaxPool1d(2, 2),
|
|
nn.Dropout(dropout_conv) # 在卷积层后添加 Dropout
|
|
)
|
|
self.CONV2 = nn.Sequential(
|
|
nn.Conv1d(64, 128, 5, 1, padding=2),
|
|
nn.BatchNorm1d(128),
|
|
nn.ReLU(),
|
|
nn.MaxPool1d(2, 2),
|
|
nn.Dropout(dropout_conv) # 在卷积层后添加 Dropout
|
|
)
|
|
self.CONV3 = nn.Sequential(
|
|
nn.Conv1d(128, 256, 3, 1, padding=1),
|
|
nn.BatchNorm1d(256),
|
|
nn.ReLU(),
|
|
nn.AdaptiveMaxPool1d(1),
|
|
nn.Dropout(dropout_conv) # 在卷积层后添加 Dropout
|
|
)
|
|
self.fc = nn.Sequential(
|
|
nn.Linear(256, 128),
|
|
nn.ReLU(),
|
|
nn.Dropout(dropout_fc), # 全连接层中的 Dropout
|
|
nn.Linear(128, nls)
|
|
)
|
|
|
|
def forward(self, x):
|
|
x = self.CONV1(x)
|
|
x = self.CONV2(x)
|
|
x = self.CONV3(x)
|
|
x = x.view(x.size(0), -1)
|
|
out = self.fc(x)
|
|
return out
|
|
|
|
|
|
# 训练函数
|
|
def CNNTrain(X_train, X_test, y_train, y_test, BATCH_SIZE, n_epochs, nls, model_path):
|
|
data_train, data_test = ZspPocess(X_train, X_test, y_train, y_test, need=True)
|
|
train_loader = torch.utils.data.DataLoader(data_train, batch_size=BATCH_SIZE, shuffle=True)
|
|
test_loader = torch.utils.data.DataLoader(data_test, batch_size=BATCH_SIZE, shuffle=False)
|
|
|
|
model = CNN3Layers(nls=nls, dropout_conv=0.3, dropout_fc=0.5).to(device)
|
|
optimizer = optim.Adam(model.parameters(), lr=0.0001, weight_decay=0.001)
|
|
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, 'min', factor=0.5, patience=5)
|
|
criterion = nn.CrossEntropyLoss().to(device)
|
|
scaler = GradScaler()
|
|
|
|
best_acc = 0.0
|
|
model_save_path = model_path
|
|
|
|
for epoch in range(n_epochs):
|
|
model.train()
|
|
train_acc, train_loss = [], []
|
|
|
|
for i, data in enumerate(train_loader):
|
|
inputs, labels = data
|
|
inputs = inputs.to(device).float()
|
|
labels = labels.to(device).long()
|
|
|
|
optimizer.zero_grad()
|
|
|
|
with autocast():
|
|
outputs = model(inputs)
|
|
loss = criterion(outputs, labels)
|
|
|
|
scaler.scale(loss).backward()
|
|
scaler.step(optimizer)
|
|
scaler.update()
|
|
|
|
_, predicted = torch.max(outputs.data, 1)
|
|
acc = accuracy_score(labels.cpu(), predicted.cpu())
|
|
train_acc.append(acc)
|
|
train_loss.append(loss.item())
|
|
|
|
avg_train_loss = np.mean(train_loss)
|
|
avg_train_acc = np.mean(train_acc)
|
|
|
|
writer.add_scalar('Loss/train', avg_train_loss, epoch)
|
|
writer.add_scalar('Accuracy/train', avg_train_acc, epoch)
|
|
|
|
# 测试集评估
|
|
model.eval()
|
|
test_acc, test_loss, test_precision, test_recall, test_f1 = [], [], [], [], []
|
|
y_true, y_pred = [], []
|
|
with torch.no_grad():
|
|
for data in test_loader:
|
|
inputs, labels = data
|
|
inputs = inputs.to(device).float()
|
|
labels = labels.to(device).long()
|
|
|
|
with autocast():
|
|
outputs = model(inputs)
|
|
loss = criterion(outputs, labels)
|
|
|
|
_, predicted = torch.max(outputs.data, 1)
|
|
acc = accuracy_score(labels.cpu(), predicted.cpu())
|
|
precision = precision_score(labels.cpu(), predicted.cpu(), average='weighted', zero_division=1)
|
|
recall = recall_score(labels.cpu(), predicted.cpu(), average='weighted', zero_division=1)
|
|
f1 = f1_score(labels.cpu(), predicted.cpu(), average='weighted', zero_division=1)
|
|
|
|
y_true.extend(labels.cpu().numpy())
|
|
y_pred.extend(predicted.cpu().numpy())
|
|
|
|
test_acc.append(acc)
|
|
test_loss.append(loss.item())
|
|
test_precision.append(precision)
|
|
test_recall.append(recall)
|
|
test_f1.append(f1)
|
|
|
|
avg_test_loss = np.mean(test_loss)
|
|
avg_test_acc = np.mean(test_acc)
|
|
avg_test_precision = np.mean(test_precision)
|
|
avg_test_recall = np.mean(test_recall)
|
|
avg_test_f1 = np.mean(test_f1)
|
|
|
|
writer.add_scalar('Loss/test', avg_test_loss, epoch)
|
|
writer.add_scalar('Accuracy/test', avg_test_acc, epoch)
|
|
writer.add_scalar('Precision/test', avg_test_precision, epoch)
|
|
writer.add_scalar('Recall/test', avg_test_recall, epoch)
|
|
writer.add_scalar('F1_Score/test', avg_test_f1, epoch)
|
|
|
|
# 打印每个 epoch 的训练和测试结果
|
|
print(f"Epoch [{epoch + 1}/{n_epochs}]")
|
|
print(f"Train Loss: {avg_train_loss:.4f}, Train Accuracy: {avg_train_acc:.4f}")
|
|
print(f"Test Loss: {avg_test_loss:.4f}, Test Accuracy: {avg_test_acc:.4f}")
|
|
print(f"Test Precision: {avg_test_precision:.4f}, Test Recall: {avg_test_recall:.4f}, Test F1: {avg_test_f1:.4f}")
|
|
|
|
if avg_test_acc > best_acc:
|
|
best_acc = avg_test_acc
|
|
torch.save(model.state_dict(), model_save_path)
|
|
|
|
scheduler.step(avg_test_loss)
|
|
|
|
return {
|
|
"accuracy": avg_test_acc,
|
|
"precision": avg_test_precision,
|
|
"recall": avg_test_recall,
|
|
"f1_score": avg_test_f1,
|
|
"confusion_matrix": confusion_matrix(y_true, y_pred)
|
|
}
|
|
|
|
# 测试函数
|
|
def CNNtest(X_test, y_test, BATCH_SIZE, nls, model_path):
|
|
# 标准化测试数据并创建 DataLoader
|
|
scaler = StandardScaler()
|
|
X_test = scaler.fit_transform(X_test) # 只对 X_test 进行标准化
|
|
X_test = torch.tensor(X_test[:, np.newaxis, :], dtype=torch.float32)
|
|
y_test = torch.tensor(y_test, dtype=torch.long)
|
|
|
|
# 创建测试数据集和 DataLoader
|
|
data_test = MyDataset(X_test, y_test, augment=False)
|
|
test_loader = torch.utils.data.DataLoader(data_test, batch_size=BATCH_SIZE, shuffle=False)
|
|
|
|
# 加载模型结构和权重
|
|
model = CNN3Layers(nls=nls).to(device)
|
|
model.load_state_dict(torch.load(model_path))
|
|
|
|
# 初始化评估指标
|
|
y_true, y_pred = [], []
|
|
|
|
# 测试过程
|
|
model.eval()
|
|
with torch.no_grad():
|
|
for inputs, labels in test_loader:
|
|
inputs, labels = inputs.to(device).float(), labels.to(device).long()
|
|
outputs = model(inputs)
|
|
_, predicted = torch.max(outputs.data, 1)
|
|
|
|
# 收集真实标签和预测标签
|
|
y_true.extend(labels.cpu().numpy())
|
|
y_pred.extend(predicted.cpu().numpy())
|
|
|
|
# 计算评估指标
|
|
accuracy = accuracy_score(y_true, y_pred)
|
|
precision = precision_score(y_true, y_pred, average='weighted')
|
|
recall = recall_score(y_true, y_pred, average='weighted')
|
|
f1 = f1_score(y_true, y_pred, average='weighted')
|
|
cm = confusion_matrix(y_true, y_pred)
|
|
|
|
# 返回评估结果
|
|
return {
|
|
"accuracy": accuracy,
|
|
"precision": precision,
|
|
"recall": recall,
|
|
"f1_score": f1,
|
|
"confusion_matrix": cm
|
|
}
|
|
|
|
|
|
def optimize_CNN(X_train, X_test, y_train, y_test, model_path):
|
|
# 贝叶斯优化的搜索空间
|
|
param_space = {
|
|
'batch_size': Integer(16, 128), # batch size 的范围
|
|
'n_epochs': Integer(10, 100), # 训练 epochs 的范围
|
|
'dropout_conv': Real(0.1, 0.5, 'uniform'), # 卷积层 dropout 比例
|
|
'dropout_fc': Real(0.1, 0.5, 'uniform'), # 全连接层 dropout 比例
|
|
'lr': Real(1e-5, 1e-2, 'log-uniform'), # 学习率范围
|
|
}
|
|
|
|
# 训练模型的目标函数
|
|
def objective(params):
|
|
batch_size, n_epochs, dropout_conv, dropout_fc, lr = params
|
|
|
|
# 使用给定的超参数进行训练
|
|
train_metrics = CNNTrain(
|
|
X_train, X_test, y_train, y_test,
|
|
BATCH_SIZE=batch_size, n_epochs=n_epochs,
|
|
nls=21, model_path=model_path,
|
|
)
|
|
|
|
# 测试模型并返回评估指标
|
|
test_metrics = CNNtest(X_test, y_test, batch_size, nls=21, model_path=model_path)
|
|
|
|
# 我们以测试集的 accuracy 作为优化目标
|
|
return -test_metrics["accuracy"] # 贝叶斯优化是最小化目标函数，所以返回负值
|
|
|
|
# 使用贝叶斯优化进行调优
|
|
optimizer = BayesSearchCV(
|
|
estimator=None, # 不使用具体的模型，这里我们将目标函数传给贝叶斯优化
|
|
search_spaces=param_space, # 搜索空间
|
|
n_iter=20, # 调优的迭代次数
|
|
n_jobs=-1, # 使用所有可用的 CPU 核心
|
|
verbose=1, # 输出优化过程
|
|
random_state=42, # 固定随机种子
|
|
)
|
|
|
|
# 进行超参数调优
|
|
optimizer.fit(X_train, y_train)
|
|
|
|
# 输出最优超参数
|
|
best_params = optimizer.best_params_
|
|
print("Best hyperparameters:", best_params)
|
|
|
|
# 使用最优超参数训练并返回评估指标
|
|
batch_size = best_params['batch_size']
|
|
n_epochs = best_params['n_epochs']
|
|
dropout_conv = best_params['dropout_conv']
|
|
dropout_fc = best_params['dropout_fc']
|
|
lr = best_params['lr']
|
|
|
|
train_metrics = CNNTrain(
|
|
X_train, X_test, y_train, y_test,
|
|
BATCH_SIZE=batch_size, n_epochs=n_epochs,
|
|
nls=21, model_path=model_path,
|
|
)
|
|
|
|
test_metrics = CNNtest(X_test, y_test, batch_size, nls=21, model_path=model_path)
|
|
|
|
# 返回训练和测试的评估结果
|
|
return best_params, train_metrics, test_metrics
|