Stopper.which_best_in_recent_history()

Stopper.which_best_in_recent_history()#

Stopper.which_best_in_recent_history(i, loss_history)[source]#

Identifies the index of the best observation in the recent loss window.

The recent loss window contains loss_history[i - patience : i], where i is the number of completed epochs. This returns the best index in that recent window, not necessarily the global best index in the full loss history.