Multiclassification: objectives and metrics¶
Objectives and metrics¶
The formulas use the variables common to all metrics.
softmax¶
\[
\displaystyle\frac{-\sum\limits_{i=1}^{N}w_{i}\log\left(\displaystyle\frac{e^{a_{it_{i}}}}{ \sum\limits_{j=0}^{M - 1}e^{a_{ij}}} \right)}{\sum\limits_{i=1}^{N}w_{i}} { ,}
\]
\(t \in \{0, ..., M - 1\}\)
where \(a_{ij}\) is the raw score of the i-th object for class \(j\).
TBoostClassifier uses this objective automatically when the target has three or more distinct
values (with objective="logistic"). The model has one set of rating tables per class: the raw
score of a class is its intercept plus its tables, and the probabilities are the softmax of the
class scores.
Note
Multiclassification does not support exposure, graduation, reanchor, reanchor_slope
or the parameters listed in Advanced settings.
Used for optimization¶
| Name | Optimization | Link |
|---|---|---|
| softmax | + | softmax |