Hacker News new | past | comments | ask | show | jobs | submit login

"DNNs are nothing more than a huge set of NNs"

As in deep neural networks are nothing more than a huge set of shallow neural networks? That's directly contradicted by tons of evidence. See for example the visualizations in https://distill.pub/2017/feature-visualization/

Your claim 2 is of course correct, supported by the phenomenon of transfer learning.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: