Skip to main content

Crate gam_solve

Crate gam_solve 

Source

Re-exports§

pub use evidence::EvidenceHvpLogDet;
pub use evidence::EvidenceIftGradientTerms;
pub use evidence::EvidenceLogDetSource;
pub use evidence::GaussianMixtureConfig;
pub use evidence::GaussianMixtureFit;
pub use evidence::SelectedTopology;
pub use evidence::StackingConfig;
pub use evidence::StackingWeights;
pub use evidence::TopologyCandidate;
pub use evidence::TopologyKind;
pub use evidence::TopologyScoreScale;
pub use evidence::TopologySelectOptions;
pub use evidence::UNION_STRUCTURE_LADDER;
pub use evidence::UnionComponentFit;
pub use evidence::UnionComponentKind;
pub use evidence::UnionStructure;
pub use evidence::UnionStructureFit;
pub use evidence::evidence_grad_rho;
pub use evidence::evidence_hessian_log_det;
pub use evidence::evidence_ift_gradient_correction;
pub use evidence::fit_gaussian_mixture;
pub use evidence::fit_union_ladder;
pub use evidence::fit_union_structure;
pub use evidence::hessian_log_det_from_hvp;
pub use evidence::laplace_evidence;
pub use evidence::select_topology;
pub use evidence::solve_stacking_weights;
pub use evidence::union_per_point_log_density;
pub use evidence::union_responsibility_split;
pub use topology_selector::AutoTopologyKind;
pub use topology_selector::CrossClassCandidate;
pub use topology_selector::CrossClassRaceVerdict;
pub use topology_selector::EvidenceCertification;
pub use topology_selector::Headline;
pub use topology_selector::HeldOutDensityProvider;
pub use topology_selector::MIXTURE_K_LADDER;
pub use topology_selector::MixtureRungFit;
pub use topology_selector::MixtureRungResult;
pub use topology_selector::STACKING_CV_FOLDS;
pub use topology_selector::STACKING_CV_SEED;
pub use topology_selector::TopologyAutoFitEvidence;
pub use topology_selector::TopologyAutoRankedFit;
pub use topology_selector::TopologyAutoSelector;
pub use topology_selector::TopologyAutoSelectorResult;
pub use topology_selector::TopologyRaceParallelCandidate;
pub use topology_selector::UnionRungFit;
pub use topology_selector::UnionRungResult;
pub use topology_selector::adjudicate_cross_class_race;
pub use topology_selector::build_cv_log_density_table;
pub use topology_selector::deterministic_cv_folds;
pub use topology_selector::deterministic_cv_folds_seeded;
pub use topology_selector::fit_mixture_rung;
pub use topology_selector::fit_union_candidate;
pub use topology_selector::fit_union_rung;
pub use topology_selector::mixture_density_provider;
pub use topology_selector::parse_union_name;
pub use topology_selector::run_topology_race_parallel;
pub use topology_selector::select_topology_with_fit;
pub use topology_selector::select_topology_with_fit_parallel;
pub use topology_selector::tk_normalized_score;
pub use topology_selector::union_density_provider;
pub use estimate::reml::reml_outer_engine::BarrierConfig;
pub use estimate::reml::reml_outer_engine::CorrectedCovarianceError;

Modules§

active_set
arrow_schur
Bundle-adjustment Schur solver for joint (t, β) inner systems.
continuation_path
Object 1 — the ContinuationPath: one object that couples the three annealing schedules that today live separately and uncoupled, so a K≥2 SAE joint fit always arrives via a regime where the inner problem is a contraction — never solved cold.
cross_node
Cross-node deterministic border-Gram reduction (#987, extending #973).
estimate
Model Estimation via Penalized Likelihood and REML
evidence
Canonical Laplace evidence, IFT cascade, and topology selection.
gauge
gaussian_reml
glm_sufficient_lane
Frozen-weight weighted-slab Gram tensor: the n-independence FIRST INSTANCE for the design-moving GLM/PIRLS ψ-sweep (#1033 mechanism (c), first step).
gpu
Solver GPU kernels and dispatch wiring.
gpu_kernels
inference
Inference-tier numerics that are genuine gam-solve criterion math (descended #1521): leaf evidence/diagnostic computations whose dependencies are all at or below the gam-solve tier (estimate, pirls, sensitivity, mixture_link, gam-linalg, gam-problem). The monolith crate root re-exports this subtree as gam::inference::*, so existing callers (gam::inference::alo, …) resolve unchanged.
inner_status
Structured return type for the custom-family inner blockwise solver.
latent_cache
Persistent cache for latent-coordinate REML design evaluations.
latent_inner
Joint (t, β) inner Newton driver for gam_terms::latent::LatentCoordValues blocks.
logdet_bounds
#1011 — deterministic two-sided enclosures for a block-SPD log-determinant.
loop_guard
Certified termination (#968): ONE exhaustion/stagnation policy for every damped inner loop.
measure_jet_glm_sufficient
GLM fixed-design sufficient-statistic reuse for #1033 mechanism (a), extending the n-free lever from the Gaussian lane to non-Gaussian families.
measure_jet_gram_cache
Sufficient-statistic caches for #1033 mechanism (a), the measure-jet fixed-design case.
mixture_link
model_types
Re-exported for the Python bindings (gam-pyffi), which build their analytic-penalty registry through the single shared descriptor parser that also serves the in-process workflow pipeline. Exposed here so PyFFI can name it without the (crate-private) workflow module being publicly reachable.
objective_base
orthogonal_reparam
General exact orthogonal reparameterization of overlapping design blocks (universal robustness — the “orthogonalize” stage).
parallel_strategy
Workload-aware parallel scheduling for row reductions.
persistent_warm_start
pirls
The penalized iteratively-reweighted-least-squares (P-IRLS) inner solver, split by concern into real submodules. Each concern module pulls the shared crate-level import surface and every sibling’s items through a single use super::*;, backed by the pub(crate) use <mod>::*; re-exports below — reproducing the flat namespace the module historically shared, without any line-cut include! fragments.
progress_log
Stderr progress logger for the gam CLI and the gamfit Python bindings.
psi_gram_tensor
Certified Chebyshev-in-ψ Gram tensor: n-independent design-moving trials (#1033 item b).
psis
Pareto-smoothed importance-sampling utilities.
quadrature
Gauss-Hermite Quadrature for Posterior Mean Predictions
residual_cascade
Multiresolution residual cascade for scattered 2-3D smooths at huge n (compute-first primitive #3, #1032; siblings: the 1-D scan in crate::spline_scan, the 2-D grid in gam_terms::grid_spline_2d).
rho_optimizer
Central authority for outer smoothing-parameter optimization strategy.
rho_prior_eval
Shared evaluation of the configured deterministic smoothing-parameter (ρ) prior objective.
rho_uncertainty
PSIS diagnostic for marginal smoothing-parameter uncertainty.
row_measure
Row-subsample mask handle for trust-region invariant enforcement.
row_sampling_measure
RowSamplingMeasure — the Fisher-mass enrichment producer (role (c) of #980).
seeding
sensitivity
ONE sensitivity operator (#935): every “how does the fit move?” question is the same solve.
spline_scan
Exact O(n) state-space polynomial smoothing spline (“the scan”).
streaming_border
Streaming, deterministic, out-of-core border-Gram accumulation (#973).
structure_search
#976 — evidence-guarded dictionary structure search: atom birth / death / fission / fusion as anytime-valid hypothesis tests, with a deterministic, serializable SearchLedger as the honesty surface.
topology_formula
Formula-string surgery for select_topology AUTO substitution.
topology_selector
Auto-selection helpers for latent-coordinate topology candidates.
warm_start_artifact
Cross-fit warm-start artifact: a descriptor-indexed, function-space snapshot of a converged fit, designed so a related later fit (a leave-one-subject-out fold, a re-fit on a different row population, a different reduced width) can warm-start from it even though the exact response-keyed inner cache (persistent_warm_start.rs) misses.
warm_start_transfer
Cross-fit warm-start transfer: build a starting ρ iterate for a new fit from a structurally-matching parent FitArtifact.

Statics§

SMOOTHING_CORRECTION_CUBATURE_COUNT
Process-wide counter of smoothing-corrections that took the sigma-cubature (second-order) branch in estimate::reml::eval::RemlState::compute_smoothing_correction_auto. Re-exported so integration tests can snapshot it before/after a fit and prove the cubature path (rather than the first-order linearization) was actually exercised — see the #582 response-scale-equivariance regression. Process-wide count of cubature upgrades that succeeded inside [RemlState::compute_smoothing_correction_auto]. Paired with [SMOOTHING_CORRECTION_NUMERICAL_FAILURE_COUNT] for visibility.