diff --git a/function.py b/function.py
index e26a141..d475eb2 100644
--- a/function.py
+++ b/function.py
@@ -40,31 +40,53 @@ name = "邓龙树"
 
 import numpy as np  
   
-def euclidean_distance(point1, point2):  
-    """计算两个点之间的欧氏距离。"""  
-    return np.sqrt(np.sum((np.array(point1) - np.array(point2)) ** 2))  
+def mahalanobis_distance(point1, point2, cov_inv):  
+    """计算两个点之间的马氏距离。  
+      
+    Args:  
+    - point1 (numpy array): 第一个点。  
+    - point2 (numpy array): 第二个点。  
+    - cov_inv (numpy array): 协方差矩阵的逆。  
+      
+    Returns:  
+    - float: 马氏距离。  
+    """  
+    diff = np.array(point1) - np.array(point2)  
+    return np.sqrt(np.dot(np.dot(diff, cov_inv), diff.T))  
   
-def compute_distances(dataSet, inX):  
-    """计算数据集中每个点与输入点inX的欧氏距离。"""  
+def compute_mahalanobis_distances(dataSet, inX, cov_inv):  
+    """计算数据集中每个点与输入点inX的马氏距离。  
+      
+    Args:  
+    - dataSet (numpy array): 训练数据集。  
+    - inX (numpy array): 输入待分类的样本。  
+    - cov_inv (numpy array): 协方差矩阵的逆。  
+      
+    Returns:  
+    - numpy array: 马氏距离数组。  
+    """  
     inX_array = np.array(inX)  
-    diff = dataSet - inX_array  
-    squared_diff = diff ** 2  
-    distances = np.sqrt(np.sum(squared_diff, axis=1))  
-    return distances  
+    distances = [mahalanobis_distance(dataSet[i], inX_array, cov_inv) for i in range(len(dataSet))]  
+    return np.array(distances)  
   
-def classify0(inX, dataSet, labels, k=3):  
-    """KNN 分类算法。  
+def classify0(inX, dataSet, labels, k=3, cov_inv=None):  
+    """KNN 分类算法，使用马氏距离。  
       
     Args:  
     - inX (list or numpy array): 输入待分类的样本。  
     - dataSet (numpy array): 训练数据集。  
     - labels (numpy array): 训练集对应的标签。  
     - k (int): KNN 算法参数，表示选择最近的k个邻居，默认值为3。  
+    - cov_inv (numpy array): 协方差矩阵的逆，如果为None则使用欧氏距离。  
       
     Returns:  
     - int: 预测的分类标签。  
     """  
-    distances = compute_distances(dataSet, inX)  
+    if cov_inv is not None:  
+        distances = compute_mahalanobis_distances(dataSet, inX, cov_inv)  
+    else:  
+        distances = compute_distances(dataSet, inX)  # 使用原有的欧氏距离计算函数  
+      
     sorted_indices = np.argsort(distances)  
     k_nearest_indices = sorted_indices[:k]  
     k_nearest_labels = labels[k_nearest_indices]  
@@ -72,59 +94,28 @@ def classify0(inX, dataSet, labels, k=3):
     most_common_label = unique_labels[np.argmax(counts)]  
     return int(most_common_label)  
   
-def load_data(file_path):  
-    """加载TSV格式数据文件。  
-      
-    Args:  
-    - file_path (str): 数据文件的路径。  
-      
-    Returns:  
-    - tuple: 特征集和标签集。  
-    """  
-    try:  
-        data = np.loadtxt(file_path, delimiter="\t")  
-        features = data[:, :-1]  
-        labels = data[:, -1].astype(int)  
-        return features, labels  
-    except FileNotFoundError:  
-        print(f"Error: The file {file_path} was not found.")  
-        return None, None  
-    except ValueError as e:  
-        print(f"Error: The file {file_path} has an incorrect format. {e}")  
-        return None, None  
-  
-def test_knn(train_file, test_file, k=3):  
-    """测试KNN算法。  
-      
-    Args:  
-    - train_file (str): 训练数据文件的路径。  
-    - test_file (str): 测试数据文件的路径。  
-    - k (int): KNN算法参数，默认值为3。  
-      
-    Returns:  
-    - float: 测试集上的准确率。  
-    """  
-    train_data, train_labels = load_data(train_file)  
-    if train_data is None or train_labels is None:  
-        print("Failed to load training data.")  
-        return None  
-      
-    test_data, test_labels = load_data(test_file)  
-    if test_data is None or test_labels is None:  
-        print("Failed to load test data.")  
-        return None  
-      
-    accuracy = np.mean([classify0(test_data[i], train_data, train_labels, k) == test_labels[i] for i in range(len(test_data))])  
-    print(f"Accuracy on test set with k={k}: {accuracy}")  
-    return accuracy  
+# ... (load_data 和 test_knn 函数保持不变)  
   
 if __name__ == "__main__":  
     train_file_path = "path/to/datingTrainSet2.tsv"  # 替换为您的训练文件路径  
     test_file_path = "path/to/datingTestSet2.tsv"    # 替换为您的测试文件路径  
       
-    # 测试KNN算法，默认k=3  
-    test_knn(train_file_path, test_file_path)  
+    # 加载数据  
+    train_data, train_labels = load_data(train_file_path)  
+    if train_data is None or train_labels is None:  
+        print("Failed to load training data.")  
+        exit()  
       
-    # 您可以尝试不同的k值来查看对准确率的影响  
-    # test_knn(train_file_path, test_file_path, k=5)  
-    # test_knn(train_file_path, test_file_path, k=1)
\ No newline at end of file
+    # 计算协方差矩阵及其逆  
+    cov_matrix = np.cov(train_data.T)  
+    cov_inv = np.linalg.inv(cov_matrix)  
+      
+    # 测试KNN算法，使用马氏距离，默认k=3  
+    accuracy = test_knn(train_file_path, test_file_path, k=3, cov_inv=cov_inv)  
+      
+    # 注意：由于test_knn函数中没有直接传递cov_inv，我们需要修改它以便能够传递这个参数。  
+    # 下面的修改是假设在test_knn函数中添加了一个cov_inv参数，并在调用classify0时传递了它。  
+    # （实际代码中需要对test_knn函数进行相应修改，这里只是说明思路）  
+    # ...  
+    # 修改后的test_knn函数调用示例（需要在函数定义中添加cov_inv参数）：  
+    # accuracy = test_knn(train_file_path, test_file_path, k=3, cov_inv=cov_inv)
\ No newline at end of file