508012 (2) [Avatar] Offline
#1
In the plot, training values should be solid lines and validation values should be dots.
svenproppert (5) [Avatar] Offline
#2
I agree.

Another more general suggestion: Should labels be introduced to the plots in order to facilitate the interpretation?
Shuyib (6) [Avatar] Offline
#3
I got the source code which is now available and run the jupyter notebook for example the imdb dataset this is what i got. I think as long as the dots and lines are labeled everything's fine. I've attached it.