The hand drawn graph shows the large NN increasing as data increased without plataeuing. I then took this information to answer one of the questions and it was marked as wrong. The question was roughly asking if NNs keep increasing in peformance as we add more data without increasing model size. The small and medium NNs plateaued so I thought this meant a large NN didn’t necessarily plateau in performance, as that’s what the graph showed.
I just think the graph needs redoing or maybe I missed something obvious?
So the lines for each NN dont all plateau. Which the question said happens as you increase the data. In the graph it show Large NN keeps increasing in performance regardless of data size. But the question/answer said this was wrong.
