to 
We are building in inductive biases into ANNs without properly understanding how they work. But it does work. Isn't this a bit like evolution?https://twitter.com/zacharylipton/status/1354649259749027841 …
-
Then it might take many millions of years of exploration to 'evolve' networks with the robustness and adaptability of biological intelligence :( ...
-
-
otoh, optimization in a computer is faster. And hey, I belive the world is spending a lot more on hyperparameter optimization than on computer per se (although that may be only true on research side). In any case, a few billion generations may happen faster than we think.
-
It's incredible how many different open-sources hyperparameter optimisation softwares exist. Should we also be investing more into AutoML? It would dramatically simplify the workflow.
- Pokaż odpowiedzi
Nowa rozmowa -
Wydaje się, że ładowanie zajmuje dużo czasu.
Twitter jest przeciążony lub wystąpił chwilowy problem. Spróbuj ponownie lub sprawdź status Twittera, aby uzyskać więcej informacji.