Názor k článku Architektura a novinky GPU Turing. Detailní pohled na novou architekturu Nvidie od Maudit - Pouzivaji se pri uceni a i k inferenci....

Pouzivaji se pri uceni a i k inferenci. Uceni pomoci backpropagation nastavuje vahu spojeni neuronu mezi vrstvami v rezimu kazdy s kazdym mezi dvema vrstvami. Ta naucena vaha kazdeho spojeni se pak pouziva k lepsimu vyhodnoceni vstupu.

Samotny proces uceni bere vstupy a prozene je siti, zpatky se pak vraci informace o tom, jak dobre/spatne se vystup trefil.

"Backpropagation is a method used in artificial neural networks to calculate a gradient that is needed in the calculation of the weights to be used in the network.[1] It is commonly used to train deep neural networks,[2] a term referring to neural networks with more than one hidden layer.[3]"
https://en.wikipedia.org/wiki/Backpropagation