pytorch中計算準確率,召回率和F1值的方法

# PyTorch中計算準確率、召回率和F1值的方法

## 1. 引言

在機器學習分類任務中，評估模型性能是至關重要的環節。準確率(Accuracy)、召回率(Recall)和F1值(F1-Score)是最常用的評估指標之一。本文將詳細介紹如何在PyTorch框架下實現這些指標的計算，包括理論基礎、實現方法和實際應用示例。

## 2. 分類任務評估指標基礎

### 2.1 混淆矩陣

混淆矩陣(Confusion Matrix)是理解分類指標的基礎。對于二分類問題，混淆矩陣如下：

|                | 預測為正類 | 預測為負類 |
|----------------|------------|------------|
| **實際為正類** | TP (真正例) | FN (假反例) |
| **實際為負類** | FP (假正例) | TN (真反例) |

### 2.2 指標定義

- **準確率(Accuracy)**: 正確預測的樣本比例
  $$Accuracy = \frac{TP + TN}{TP + TN + FP + FN}$$

- **召回率(Recall)**: 正類樣本中被正確預測的比例
  $$Recall = \frac{TP}{TP + FN}$$

- **精確率(Precision)**: 預測為正類的樣本中實際為正類的比例
  $$Precision = \frac{TP}{TP + FP}$$

- **F1值(F1-Score)**: 精確率和召回率的調和平均
  $$F1 = 2 \times \frac{Precision \times Recall}{Precision + Recall}$$

## 3. PyTorch實現基礎

在PyTorch中計算這些指標，我們需要處理模型的輸出和真實標簽。通常分類模型的輸出是每個類別的概率(logits)，我們需要先將其轉換為預測類別。

### 3.1 獲取預測結果

```python
import torch

# 假設模型輸出為logits (batch_size × num_classes)
logits = torch.randn(4, 3)  # 4個樣本，3分類問題

# 獲取預測類別
_, preds = torch.max(logits, dim=1)  # 獲取每行最大值的索引

3.2 處理真實標簽

真實標簽通常是類別索引形式：

targets = torch.tensor([0, 2, 1, 1])  # 真實標簽

4. 二分類指標實現

4.1 準確率計算

def accuracy_binary(preds, targets):
    correct = (preds == targets).float()
    acc = correct.mean()
    return acc

4.2 召回率和精確率

def precision_recall_binary(preds, targets, positive_class=1):
    true_positives = ((preds == positive_class) & (targets == positive_class)).sum().float()
    predicted_positives = (preds == positive_class).sum().float()
    actual_positives = (targets == positive_class).sum().float()
    
    precision = true_positives / (predicted_positives + 1e-8)  # 避免除以0
    recall = true_positives / (actual_positives + 1e-8)
    
    return precision, recall

4.3 F1值計算

def f1_score_binary(preds, targets, positive_class=1):
    precision, recall = precision_recall_binary(preds, targets, positive_class)
    f1 = 2 * (precision * recall) / (precision + recall + 1e-8)
    return f1

5. 多分類指標實現

對于多分類問題，我們有兩種計算方式： - 宏平均(Macro-average): 對每個類別的指標單獨計算后平均 - 微平均(Micro-average): 將所有類別的TP,FP,FN等先求和再計算

5.1 宏平均實現

def macro_precision_recall_f1(preds, targets, num_classes):
    # 初始化統計量
    class_stats = []
    
    for class_idx in range(num_classes):
        # 計算當前類別的TP, FP, FN
        tp = ((preds == class_idx) & (targets == class_idx)).sum().float()
        fp = ((preds == class_idx) & (targets != class_idx)).sum().float()
        fn = ((preds != class_idx) & (targets == class_idx)).sum().float()
        
        precision = tp / (tp + fp + 1e-8)
        recall = tp / (tp + fn + 1e-8)
        f1 = 2 * (precision * recall) / (precision + recall + 1e-8)
        
        class_stats.append((precision, recall, f1))
    
    # 計算宏平均
    macro_precision = torch.mean(torch.tensor([s[0] for s in class_stats]))
    macro_recall = torch.mean(torch.tensor([s[1] for s in class_stats]))
    macro_f1 = torch.mean(torch.tensor([s[2] for s in class_stats]))
    
    return macro_precision, macro_recall, macro_f1

5.2 微平均實現

def micro_precision_recall_f1(preds, targets, num_classes):
    # 初始化全局統計量
    total_tp = 0
    total_fp = 0
    total_fn = 0
    
    for class_idx in range(num_classes):
        tp = ((preds == class_idx) & (targets == class_idx)).sum().float()
        fp = ((preds == class_idx) & (targets != class_idx)).sum().float()
        fn = ((preds != class_idx) & (targets == class_idx)).sum().float()
        
        total_tp += tp
        total_fp += fp
        total_fn += fn
    
    micro_precision = total_tp / (total_tp + total_fp + 1e-8)
    micro_recall = total_tp / (total_tp + total_fn + 1e-8)
    micro_f1 = 2 * (micro_precision * micro_recall) / (micro_precision + micro_recall + 1e-8)
    
    return micro_precision, micro_recall, micro_f1

6. 使用PyTorch內置函數

PyTorch提供了一些內置函數可以簡化計算：

6.1 準確率計算

def accuracy_torch(preds, targets):
    return (preds == targets).float().mean()

6.2 混淆矩陣

from sklearn.metrics import confusion_matrix
import numpy as np

def get_confusion_matrix(preds, targets, num_classes):
    preds_np = preds.cpu().numpy()
    targets_np = targets.cpu().numpy()
    return confusion_matrix(targets_np, preds_np, labels=list(range(num_classes)))

7. 實際應用示例

7.1 訓練循環中的指標計算

def train_epoch(model, dataloader, criterion, optimizer, device, num_classes):
    model.train()
    total_loss = 0
    total_acc = 0
    total_precision = 0
    total_recall = 0
    total_f1 = 0
    num_batches = len(dataloader)
    
    for inputs, targets in dataloader:
        inputs, targets = inputs.to(device), targets.to(device)
        
        # 前向傳播
        outputs = model(inputs)
        _, preds = torch.max(outputs, 1)
        
        # 計算損失
        loss = criterion(outputs, targets)
        
        # 反向傳播和優化
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        
        # 計算指標
        batch_acc = accuracy_torch(preds, targets)
        batch_precision, batch_recall, batch_f1 = macro_precision_recall_f1(preds, targets, num_classes)
        
        # 累計統計
        total_loss += loss.item()
        total_acc += batch_acc.item()
        total_precision += batch_precision.item()
        total_recall += batch_recall.item()
        total_f1 += batch_f1.item()
    
    # 計算平均指標
    avg_loss = total_loss / num_batches
    avg_acc = total_acc / num_batches
    avg_precision = total_precision / num_batches
    avg_recall = total_recall / num_batches
    avg_f1 = total_f1 / num_batches
    
    return avg_loss, avg_acc, avg_precision, avg_recall, avg_f1

7.2 驗證/測試循環

def evaluate(model, dataloader, criterion, device, num_classes):
    model.eval()
    total_loss = 0
    total_acc = 0
    total_precision = 0
    total_recall = 0
    total_f1 = 0
    num_batches = len(dataloader)
    
    with torch.no_grad():
        for inputs, targets in dataloader:
            inputs, targets = inputs.to(device), targets.to(device)
            
            # 前向傳播
            outputs = model(inputs)
            _, preds = torch.max(outputs, 1)
            
            # 計算損失
            loss = criterion(outputs, targets)
            
            # 計算指標
            batch_acc = accuracy_torch(preds, targets)
            batch_precision, batch_recall, batch_f1 = macro_precision_recall_f1(preds, targets, num_classes)
            
            # 累計統計
            total_loss += loss.item()
            total_acc += batch_acc.item()
            total_precision += batch_precision.item()
            total_recall += batch_recall.item()
            total_f1 += batch_f1.item()
    
    # 計算平均指標
    avg_loss = total_loss / num_batches
    avg_acc = total_acc / num_batches
    avg_precision = total_precision / num_batches
    avg_recall = total_recall / num_batches
    avg_f1 = total_f1 / num_batches
    
    return avg_loss, avg_acc, avg_precision, avg_recall, avg_f1

8. 高級話題：多標簽分類的指標計算

對于多標簽分類(一個樣本可以屬于多個類別)，指標計算有所不同：

def multilabel_metrics(preds, targets, threshold=0.5):
    # 假設preds是sigmoid后的概率
    preds_binary = (preds > threshold).float()
    
    # 計算TP, FP, FN
    tp = (preds_binary * targets).sum(dim=0)
    fp = (preds_binary * (1 - targets)).sum(dim=0)
    fn = ((1 - preds_binary) * targets).sum(dim=0)
    
    # 計算各指標
    precision = tp / (tp + fp + 1e-8)
    recall = tp / (tp + fn + 1e-8)
    f1 = 2 * (precision * recall) / (precision + recall + 1e-8)
    
    # 微平均
    micro_precision = tp.sum() / (tp.sum() + fp.sum() + 1e-8)
    micro_recall = tp.sum() / (tp.sum() + fn.sum() + 1e-8)
    micro_f1 = 2 * (micro_precision * micro_recall) / (micro_precision + micro_recall + 1e-8)
    
    return {
        'precision': precision,
        'recall': recall,
        'f1': f1,
        'micro_precision': micro_precision,
        'micro_recall': micro_recall,
        'micro_f1': micro_f1
    }

9. 性能優化技巧

1. 批量計算: 盡量使用矩陣運算而非循環
2. GPU加速: 確保計算在GPU上進行
3. 內存效率: 避免不必要的中間變量
4. 使用半精度: 對于大型模型可考慮使用FP16

10. 常見問題與解決方案

10.1 類別不平衡問題

當數據集中各類別樣本數量差異很大時，準確率可能不是最佳指標。解決方案： - 使用加權指標 - 關注F1值而非準確率 - 使用過采樣/欠采樣技術

10.2 多分類閾值選擇

對于概率輸出，如何選擇最佳閾值： - 使用ROC曲線尋找最佳平衡點 - 根據業務需求調整(如醫療診斷可能更重視召回率)

10.3 指標波動問題

訓練過程中指標波動大可能原因： - 學習率設置不當 - 批量大小太小 - 數據預處理不一致

11. 總結

本文詳細介紹了在PyTorch中計算準確率、召回率和F1值的方法，包括： - 二分類和多分類場景 - 宏平均和微平均策略 - 訓練循環中的集成方法 - 多標簽分類的特殊處理 - 性能優化和常見問題解決方案

正確計算和解讀這些指標對于模型開發和評估至關重要。希望本文能為您的PyTorch項目提供有價值的參考。

12. 延伸閱讀

1. PyTorch官方文檔: https://pytorch.org/docs/stable/index.html
2. Scikit-learn指標文檔: https://scikit-learn.org/stable/modules/classes.html#module-sklearn.metrics
3. 《深度學習》- Ian Goodfellow等
4. 《機器學習實戰》- Peter Harrington

”`