cbx.utils.resampling.loss_update_resampling#
- class cbx.utils.resampling.loss_update_resampling(wait_thresh=5)[source]#
Bases:
object
Resampling based on loss update difference
- Parameters:
M (int) – The number of runs in the dynamic object the resampling is applied to.
wait_thresh (int) – The number of iterations to wait before resampling. The default is 5. If the best loss is not updated after the specified number of iterations, the ensemble is resampled.
- Return type:
The indices of the runs to resample as a numpy array.