上传新文件

logistic_regression.py

+import numpy as np
+import random
+
+def one_hot(y):
+    # 样本数量
+    m = len(y)
+    # 标签类别数量
+    n = len(set(y))
+    # one-hot编码
+    result = np.zeros((m, n))
+    for i in range(m):
+        result[i][y[i]] = 1
+    return result
+
+class Logistic_Regression:
+    def __init__(self, epochs=10, lr=0.01) -> None:
+        self.epochs = epochs
+        self.lr = lr
+    
+    def softmax(self, X, w):
+        f = np.exp(X.dot(w))
+        s = f.sum(axis=1).reshape(-1, 1)
+        return f / s
+    
+    # 用随机梯度下降法训练
+    def fit(self, X, y):
+        y_proba = one_hot(y)
+        m, n = X.shape
+        k = len(set(y))
+        w = np.zeros((n, k))
+        for _ in range(self.epochs):
+            for i in range(m):
+                idx = np.random.randint(0, m)
+                x = X[idx].reshape(1, -1)
+                gradient = x.T.dot(self.softmax(x, w) - y_proba[idx])
+                w -= self.lr * gradient
+        self.w = w
+    
+    # 预测值（概率形式）
+    def predict_proba(self, X):
+        return self.softmax(X, self.w)
+    
+    # 预测值（标签）
+    def predict(self, X):
+        return self.predict_proba(X).argmax(axis=1)
\ No newline at end of file