AI Dynamics

Global AI News Aggregator

Training Deep Networks Without Normalization Using Parameterized Tanh

By

@ylecun

–

New paper: turns out you can train deep nets without normalization layers by replacing them with a parameterized tanh()

→ View original post on X — @ylecun,

Commentaires

Leave a Reply

Your email address will not be published. Required fields are marked *

MORE ARTICLES