Astral Codex Ten Podcast cover image

The Canal Papers

Astral Codex Ten Podcast

00:00

How to Overfit a Neural Network

Overfitting and underfitting are two dichotomies that AI researchers think about a lot. A maximally overfitted model might only classify these exact four images, pixel by pixel, as dogs. In less severe cases it might only classify dogs of these four breeds or dogs photoshopped in these four positions. There's no perfect way to get the exact right fit - you just have to strike a balance.

Play episode from 18:03
Transcript

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app