sklearn.covariance（协方差）模块结构及用法

许多统计学问题需要做总体协方差矩阵评估，可以看作是数据集散点图的评估。

大多数情况下，这样一个评估会在与一个对估计质量有很大影响（大小、结构、同质性）的样本上完成。

sklearn.covariance包为在各种条件下精确评估总体协方差矩阵提供了有效工具。

我们假设观察值是独立且同分布的。

模块提供了以下四种常见估计方法：

1. Empirical covariance经验协方差 官方文档

class sklearn.covariance.EmpiricalCovariance(store_precision=True, assume_centered=False)

2. Shrunk Covariance缩减协方差

3. Sparse inverse covariance稀疏逆协方差

4. Robust Covariance Estimation 鲁棒协方差估计

• 模块内各种类及函数

covariance.EmpiricalCovariance([…])	Maximum likelihood covariance estimator
covariance.EllipticEnvelope([…])	An object for detecting outliers in a Gaussian distributed dataset.
covariance.GraphicalLasso([alpha, mode, …])	Sparse inverse covariance estimation with an l1-penalized estimator.
covariance.GraphicalLassoCV([alphas, …])	Sparse inverse covariance w/ cross-validated choice of the l1 penalty.
covariance.LedoitWolf([store_precision, …])	LedoitWolf Estimator
covariance.MinCovDet([store_precision, …])	Minimum Covariance Determinant (MCD): robust estimator of covariance.
covariance.OAS([store_precision, …])	Oracle Approximating Shrinkage Estimator
covariance.ShrunkCovariance([…])	Covariance estimator with shrinkage

covariance.empirical_covariance(X[, …])	Computes the Maximum likelihood covariance estimator
covariance.graphical_lasso(emp_cov, alpha[, …])	l1-penalized covariance estimator
covariance.ledoit_wolf(X[, assume_centered, …])	Estimates the shrunk Ledoit-Wolf covariance matrix.
covariance.oas(X[, assume_centered])	Estimate covariance with the Oracle Approximating Shrinkage algorithm.
covariance.shrunk_covariance(emp_cov[, …])	Calculates a covariance matrix shrunk on the diagonal

• Reference

1. Covariance estimation
2. sklearn中文手册
3. 协方差详解角度：历史、数学、物理意义