Struct MultinomialFitRequest 

pub struct MultinomialFitRequest<'a> {
    pub data: &'a EncodedDataset,
    pub formula: &'a str,
    pub config: &'a FitConfig,
    pub init_lambda: f64,
    pub max_iter: usize,
    pub tol: f64,
}

Canonical typed inputs for the formula-driven multinomial fit (fit_penalized_multinomial_formula).

Every frontend (Rust, CLI, Python FFI) builds this one request, so the warm-start / outer-search defaults live here rather than being duplicated per caller. config is the same canonical FitConfig the scalar formula families consume: weight_column is resolved against the dataset and honored as per-row case weights, and fields the softmax family cannot consume (offsets, noise/log-slope formulas, manual Firth, frailty, …) are rejected with a typed error instead of being silently dropped.

Fields

data: &'a EncodedDataset

formula: &'a str

config: &'a FitConfig

init_lambda: f64

Warm-start seed for every per-(class, term) smoothing parameter; λ is REML/LAML-selected, so this only seeds the outer search.

max_iter: usize

OUTER REML/LAML smoothing-parameter iteration budget.

tol: f64

Requested accuracy; drives the inner joint-Newton KKT target (see the control-split note inside the fit).

Implementations

impl<'a> MultinomialFitRequest<'a>

pub fn new( data: &'a EncodedDataset, formula: &'a str, config: &'a FitConfig, ) -> Self

The canonical production controls shared by the CLI and the Python FFI.

Trait Implementations

impl<'a> Clone for MultinomialFitRequest<'a>

fn clone(&self) -> MultinomialFitRequest<'a>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<'a> Copy for MultinomialFitRequest<'a>