Experimental Results with Default Settings
maximlf opened this issue · comments
maximlf commented
Hi, Thanks for the implementations, I just discovered this repo., very interesting.
I tried running the train.py with default settings, The results shown is bit far from what claimed on the paper (0.4928). I understand this is an @unofficial@ implementation of the approach, I will dig into the code, before that, could you share some of your thoughts on what could be improved in order to increase the accuracy? Thanks again for the work!
Maxim