疏锦行Python打卡 DAY 20 奇异值SVD分解

import numpy as np

# 创建一个矩阵 A (5x3)
A = np.array([[1, 2, 3],
              [4, 5, 6],
              [7, 8, 9],
              [10, 11, 12],
              [13, 14, 15]])
print("原始矩阵 A:")
print(A)

# 进行 SVD 分解
U, sigma, Vt = np.linalg.svd(A, full_matrices=False)
print("\n奇异值 sigma:")
print(sigma)

# 保留前 k=1 个奇异值进行降维
k = 1
U_k = U[:, :k]  # 取 U 的前 k 列，因为要保持行数不变
sigma_k = sigma[:k]  # 取前 k 个奇异值
Vt_k = Vt[:k, :]  # 取 Vt 的前 k 行，因为要保持列数不变

# 近似重构矩阵 A,常用于信号or图像筛除噪声
A_approx = U_k @ np.diag(sigma_k) @ Vt_k
print("\n保留前", k, "个奇异值后的近似矩阵 A_approx:")
print(A_approx)

# 计算近似误差
error = np.linalg.norm(A - A_approx, 'fro') / np.linalg.norm(A, 'fro')
print("\n近似误差 (Frobenius 范数相对误差):", error)


import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score

# 设置随机种子以便结果可重复
np.random.seed(42)

# 模拟数据：1000 个样本，50 个特征
n_samples = 1000
n_features = 50
X = np.random.randn(n_samples, n_features) * 10  # 随机生成特征数据
y = (X[:, 0] + X[:, 1] > 0).astype(int)  # 模拟二分类标签

# 划分训练集和测试集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
print(f"训练集形状: {X_train.shape}")
print(f"测试集形状: {X_test.shape}")

# 对训练集进行 SVD 分解
U_train, sigma_train, Vt_train = np.linalg.svd(X_train, full_matrices=False)
print(f"Vt_train 矩阵形状: {Vt_train.shape}")

# 选择保留的奇异值数量 k
k = 10
Vt_k = Vt_train[:k, :]  # 保留前 k 行，形状为 (k, 50)
print(f"保留 k={k} 后的 Vt_k 矩阵形状: {Vt_k.shape}")

# 降维训练集：X_train_reduced = X_train @ Vt_k.T
X_train_reduced = X_train @ Vt_k.T
print(f"降维后训练集形状: {X_train_reduced.shape}")

# 使用相同的 Vt_k 对测试集进行降维：X_test_reduced = X_test @ Vt_k.T
X_test_reduced = X_test @ Vt_k.T
print(f"降维后测试集形状: {X_test_reduced.shape}")

# 训练模型（以逻辑回归为例）
model = LogisticRegression(random_state=42)
model.fit(X_train_reduced, y_train)

# 预测并评估
y_pred = model.predict(X_test_reduced)
accuracy = accuracy_score(y_test, y_pred)
print(f"测试集准确率: {accuracy}")

# 计算训练集的近似误差（可选，仅用于评估降维效果）
X_train_approx = U_train[:, :k] @ np.diag(sigma_train[:k]) @ Vt_k
error = np.linalg.norm(X_train - X_train_approx, 'fro') / np.linalg.norm(X_train, 'fro')
print(f"训练集近似误差 (Frobenius 范数相对误差): {error}")

打卡：@浙大疏锦行