LedoitWolf
Shrinkage covariance with the Ledoit-Wolf data-driven shrinkage intensity — better conditioned than the sample covariance, especially when p is comparable to n.
Algorithm
Shrinks the sample covariance toward a scaled identity by the closed-form Ledoit-Wolf coefficient that minimises expected Frobenius error.
Constructor
Skigen::LedoitWolf<Scalar> model(bool assume_centered = false);
Parameters
| Parameter | Default | Description |
|---|---|---|
assume_centered | false | Skip mean subtraction. |
Methods
| Method | Description |
|---|---|
fit(X) | Estimate the shrunk covariance. |
score(X) | Gaussian log-likelihood. |
Fitted Attributes
| Accessor | Description |
|---|---|
covariance() | Shrunk covariance. |
shrinkage() | Estimated shrinkage intensity. |
Example
Skigen::LedoitWolf<double> lw;
lw.fit(X);
auto C = lw.covariance();
Verified against scikit-learn
This estimator is checked by the parity suite. See the generator tests/parity/generate_covariance_reference.py and the reference fixtures in tests/parity/data/ledoit_wolf/, exercised by tests/parity/parity_covariance.cpp.
API Reference
For full signatures see the LedoitWolf API Reference.