mx.callback.early.stop
¶
Description¶
Early stop with different conditions.
Early stopping applying different conditions: hard thresholds or epochs number from the best score. Tested with “epoch.end.callback” function.
Usage¶
mx.callback.early.stop(
train.metric = NULL,
eval.metric = NULL,
bad.steps = NULL,
maximize = FALSE,
verbose = FALSE
)
Arguments¶
Argument |
Description |
---|---|
|
Numeric. Hard threshold for the metric of the training data set (optional) |
|
Numeric. Hard threshold for the metric of the evaluating data set (if set, optional) |
|
Integer. How much epochs should gone from the best score? Use this option with evaluation data set |
|
Logical. Do your model use maximizing or minimizing optimization? |
|
Logical |