nnpiv.diagnostics.relative_wellposedness_sieve_diagnostic

nnpiv.diagnostics.relative_wellposedness_sieve_diagnostic(A, C, C_prime, *, sieve_grid=None, eta_grid=None, enforce_nested_rff=True, stability_growth_tol=5.0, **kwargs)[source]

Run the pre-estimation diagnostic over a sieve and eta path.

Parameters: A, C, C_prime : array-like

Inputs passed to relative_wellposedness_diagnostic(). Arrays must be row-aligned with common sample size n.

sieve_griditerable, optional

Sieve values. For feature_map='rff', entries map to n_features. For feature_map='polynomial', entries map to poly_degree. Entries may also be dictionaries of per-call keyword overrides. If None, defaults are chosen by feature map.

eta_gridfloat or iterable of float, optional

Stabilization values. If None, uses eta from kwargs (or 1e-6).

enforce_nested_rffbool, default=True

If True and applicable, compute one large RFF map and reuse nested prefixes across sieve values for speed and consistency.

stability_growth_tolfloat, default=5.0

Growth threshold used in the summary stable_flag calculation.

**kwargs

Additional arguments forwarded to relative_wellposedness_diagnostic().

Returns

Dictionary with two keys:

  • rows: list of row-level diagnostics for each (eta, sieve) pair.

  • summary: aggregate stability metrics and path metadata.

Return type

dict

Notes

This function is a path wrapper around relative_wellposedness_diagnostic(); it does not alter the underlying diagnostic definition.