You are viewing a single comment's thread from:
RE: Running Neural Networks on Different Frameworks - [Comparison]
The reported accuracy metrics seem very low, and the training times seem too fast, in comparison with what I recall from recent research papers. Do you really think this comes from a good test approach?
I am not sure, though if some really wants to double check, they can test and see if they are reproducible.
True. I just wanted to give you and the community a heads-up since the numbers seem inconsistent with other research.