Shrinkage

shrinkage.HybridFit

Shrinkage for Python · python · 0.2.0

Bases: shrinkage.ProximalFit

HybridFit(_native: shrinkage._shrinkage._ModelFit)

A hybrid result with optimization-scale clusters and inner/outer work.

Clustered lasso groups signed values in increasing order. Chain fused lasso groups contiguous segments in feature order, retaining cuts at zero weights. SLOPE groups decreasing nonnegative magnitudes and retains each feature's sign. Training scales can change equalities among the reported original-scale coefficients.

cluster_labels

Source: packages/python/src/shrinkage/fit.py

cluster_labels(self) -> numpy.typing.NDArray[numpy.intp]

Zero-based indices into cluster_values, shape (n_features,).

cluster_values

Source: packages/python/src/shrinkage/fit.py

cluster_values(self) -> FloatArray

Optimization-scale block values, shape (n_clusters,).

coefficient_signs

Source: packages/python/src/shrinkage/fit.py

coefficient_signs(self) -> FloatArray

Optimization-scale feature signs (-1, 0, or 1), shape (n_features,).

pg_steps

Source: packages/python/src/shrinkage/fit.py

pg_steps(self) -> int

Committed PG transitions, excluding diagnostic proximal calls.

coordinate_sweeps

Source: packages/python/src/shrinkage/fit.py

coordinate_sweeps(self) -> int

Complete block-coordinate sweeps over the current working set.

diagnostic_checks

Source: packages/python/src/shrinkage/fit.py

diagnostic_checks(self) -> int

Refreshed inner checks, including quadratic checks in logistic fits.

backtracks

Source: packages/python/src/shrinkage/fit.py

backtracks(self) -> int

Rejected PG trials followed by curvature doubling.

outer_iterations

Source: packages/python/src/shrinkage/fit.py

outer_iterations(self) -> int

Accepted logistic IRLS steps; zero for Gaussian fits.

outer_backtracks

Source: packages/python/src/shrinkage/fit.py

outer_backtracks(self) -> int

Rejected logistic trials followed by step halving; zero for Gaussian.

step_policy

Source: packages/python/src/shrinkage/fit.py

step_policy(self) -> shrinkage.HybridStep

Configured Gaussian or logistic quadratic curvature policy.

working_set_diagnostics

Source: packages/python/src/shrinkage/fit.py

working_set_diagnostics(self) -> shrinkage.WorkingSetDiagnostics

Effective working-set policy and feature and block work counters.

Inherited members