huge_tiger

Usage

huge_tiger(
    x,
    lambda_=None,
    nlambda=None,
    lambda_min_ratio=None,
    sym="or",
    verbose=True,
    backend="native",
    *,
    input_type="auto",
) -> HugeResult

Description

Convenience wrapper for huge(..., method="tiger"). The native C++ core centers and normalizes raw observations, builds the correlation matrix, generates the default lambda path when needed, and solves the correlation-domain square-root Lasso. An explicit lambda_ must be a numeric scalar, NumPy 0-D value, or non-empty one-dimensional sequence; multidimensional inputs are rejected. Values are preserved and must be finite, strictly positive, and non-increasing; ties are allowed. Generated paths contain only the longest prefix that passes the native KKT and numerical-degeneracy checks; truncation emits a RuntimeWarning. A user-supplied lambda that cannot be certified raises PyHugeError.

input_type="auto" preserves symmetry-based detection. Use "data" for square symmetric observations and "covariance" to require covariance or correlation input. After routing, covariance validation, correlation construction, automatic lambda selection, and fitting all occur in C++.