Learning Neural Network Subspaces
Mitchell Wortsman 1 Maxwell Horton 2 Carlos Guestrin 2 Ali Farhadi 2 Mohammad Rastegari 2
Abstract
Recent observations have advanced our under-
standing of the neural network optimization land-
scape, revealing the existence of (1) paths of
high accuracy containing diverse solutions and
(2) wider minima offering improved performance.
Previous methods observing diverse paths require
multiple training runs. In contrast we aim ...


雷达卡




京公网安备 11010802022788号







