Inversion types

The inverse solver returns arrays by default and can optionally return an InverseResult containing structured failures and resource statistics.

class mftt.InverseOptions(max_bracket_expansions=60, max_standardized_abs=None, max_evaluations=None, max_batch_seconds=None)[source]

Bases: object

Optional resource limits for monotone inverse workloads.

Parameters:

  • max_bracket_expansions (int, default=60) – Maximum geometric expansions used to bracket one scalar inverse root.

  • max_standardized_abs (float, optional) – Maximum absolute standardized coordinate considered during inversion.

  • max_evaluations (int, optional) – Per-row limit on component-function evaluations.

  • max_batch_seconds (float, optional) – Wall-clock limit for a row batch.

Notes

Limits turn difficult root solves into structured diagnostics rather than silently returning inaccurate values.

max_batch_seconds: float | None = None
max_bracket_expansions: int = 60
max_evaluations: int | None = None
max_standardized_abs: float | None = None
class mftt.InverseFailure(row, component, reason, residual, bracket_radius, evaluations, stage=None)[source]

Bases: object

Describe one failed inverse component solve.

Parameters:

  • row (int) – Zero-based row index in the requested inverse batch.

  • component (int) – Zero-based triangular component index that failed.

  • reason (str) – Machine-readable failure category.

  • residual (float) – Final absolute scalar residual.

  • bracket_radius (float) – Final standardized bracketing radius.

  • evaluations (int) – Number of component evaluations consumed for the row.

  • stage (int, optional) – Hierarchical fidelity-stage index, when applicable.

bracket_radius: float
component: int
evaluations: int
reason: str
residual: float
row: int
stage: int | None = None
class mftt.InverseDiagnostics(failed_mask, failures, component_failure_counts, max_residual, evaluation_counts=None)[source]

Bases: object

Aggregate inverse diagnostics for a full map inverse call.

Parameters:

  • failed_mask (numpy.ndarray, shape (n_samples,)) – Boolean mask identifying rows with at least one failed component.

  • failures (tuple of InverseFailure) – Detailed failure records.

  • component_failure_counts (numpy.ndarray, shape (input_dim,)) – Number of failed rows for each triangular component.

  • max_residual (float) – Largest finite residual, or infinity when no finite residual exists.

  • evaluation_counts (numpy.ndarray, shape (n_samples,), optional) – Total component evaluations consumed for each row.

component_failure_counts: ndarray
evaluation_counts: ndarray | None = None
failed_mask: ndarray
failures: tuple[InverseFailure, ...]
max_residual: float
class mftt.InverseResult(values, diagnostics)[source]

Bases: object

Return inverse values together with diagnostics.

Parameters:

  • values (numpy.ndarray, shape (n_samples, input_dim)) – Recovered target-space coordinates; failed rows may be incomplete.

  • diagnostics (InverseDiagnostics) – Per-row and per-component inversion outcome information.

diagnostics: InverseDiagnostics
values: ndarray