Module multiclass 

Available on crate feature alloc only.

Multi-class classification via one-vs-rest SGBT committees.

For C classes, maintains C independent SGBT ensembles, each trained with logistic loss on a binarized target (1.0 for the class, 0.0 otherwise). Final predictions are normalized via softmax across committee outputs.

Structs

MulticlassSGBT

Multi-class SGBT using one-vs-rest committee of ensembles.