更新 function.py

function.py

@@ -26,11 +26,11 @@ name = "熊浩燃"
 def sigmoid(z):
     return 1 / (1 + np.exp(-z))
 
-def min_max_normalize(X):
-    """ 数据归一化：使用Min-Max Scaling """
-    return (X - np.min(X, axis=0)) / (np.max(X, axis=0) - np.min(X, axis=0))
+def standardize(X):
+    """ 数据标准化：均值为0，标准差为1 """
+    return (X - np.mean(X, axis=0)) / (np.std(X, axis=0) + 1e-8)  # 避免除零
 
-def train_logreg(X, y, learning_rate=0.001, iterations=2000, batch_size=32, lambda_=0.1):
+def train_logreg(X, y, learning_rate=0.0005, iterations=3000, batch_size=32, lambda_=0.05):
     """
     对率回归训练函数，包含批量梯度下降和L2正则化。
 
@@ -45,9 +45,9 @@ def train_logreg(X, y, learning_rate=0.001, iterations=2000, batch_size=32, lamb
     Returns:
     - numpy array: 训练得到的权重。
     """
-    X = min_max_normalize(X)  # 使用Min-Max归一化数据
+    X = standardize(X)  # 使用标准化数据
     m, n = X.shape
-    weights = np.zeros(n)
+    weights = np.random.rand(n) * 0.01  # 初始化权重为小的随机数
     
     for _ in range(iterations):
         # 确保 batch_size 不超过 m
@@ -107,8 +107,4 @@ if __name__ == "__main__":
     X_test, y_test = load_data("./data/horse-colic.test")
 
     # 训练对率回归模型
-    weights = train_logreg(X_train, y_train, learning_rate=0.001, iterations=2000, batch_size=32, lambda_=0.1)
-
-    # 测试单个数据
-    y_pred = predict_logreg(X_test[11], weights)
-    print(f"预测结果: {y_pred}, 实际结果: {y_test[11]}")
+    weights = train_logreg(