Open andrasfuchs opened 2 years ago
@luisquintanilla
Hi @andrasfuchs
Thanks for that feedback. Providing more advanced training options is an area we have been hearing similar feedback on. Although we are not currently in the process of implementing this functionality, we'll use your suggestions as we plan for this.
I see you opened another issue that is related (#2041). Tracking this issue as part of epic #1591
Is your feature request related to a problem? Please describe. There is only one visible parameter for the train process now, the
time to train
value.Describe the solution you'd like Now that we have many good quality indicators like
Accuracy
,AUC
,AUPRC
,F1-score
for the trained models, it would be great to have a target quality score for the training, instead of just the maximum training time. After this addition the user could define what his/her target - let's say - F1-score is, so the training could stop before thetime to train
timer expiration if that quality score is achieved.Describe alternatives you've considered I can now manually stop the training if I notice that the trained model is sufficient.