> if you have enough lines you can approximate any curve within a margin How avo... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		ngrilly on Oct 20, 2020 \| parent \| context \| favorite \| on: Why deep learning works even though it shouldn’t > if you have enough lines you can approximate any curve within a margin How avoid overfitting the training data?

RandyRanderson on Oct 21, 2020 [–]

You're right - a LOT of work goes into this (eg GANs) because these large nNets just start "remembering" the training data.

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