Hello,
I am currently running an experiments and for every run, the accuracy is a bit different ~0.5.
For a paper, should one report the best one, the average or should I tweak the hyperparameters more until the difference is smaller?
Thank you
[link] [comments]