Good points you posted. I mostly agree though I never participated in such event. One thing is that recently some new approaches for training models in the sam

Good points you posted. I mostly agree though I never participated in such event. One thing is that recently some new approaches for training models in the same architectures were uncovered and are much more efficient. We may still some improvements over the same architectures but latest models showed that there always might a new best one around the corner.