def AgentSelectorWorker.fit(self, inputs, fx_choice=REG.AIG, **params) [source]
Execute the full champion selection pipeline.
Evaluates all agents on the provided data, computes the
diagnostic bundle (performance, SHAP drift, Gini path,
permutation importance), fits the champion agent, and
produces the complete audit package.
Parameters
inputs : AgentSelectorInput
Input data (OOT/test set) and the champion predictor
reference.
reference.
fx_choice : str
Fitter choice key.
**params : dict
Runtime parameter overrides, passed to the internal fitter
(select_alpha, select_bins, select_choice).
(select_alpha, select_bins, select_choice).
Returns
AgentSelector
The fitted selector instance holding the promoted champion agent.
Champion Fit: The champion is fitted via
AgentPredictor — it re-fits the model on the selected
feature subset and binds the preprocessing encoders.
The result is a self-contained deployment artifact.