发帖
雷达卡
The Heavy-Tail Phenomenon in SGD
Mert Gürbüzbalaban 1 Umut Simsekli 2 Lingjiong Zhu 3
Abstract 1. Introduction
The learning problem in neural networks can be expressed as
In recent years, various notions of capacity and an instance of the well-known population risk minimization
complexity have been proposed for characterizing problem in statistics, given as follows:
the generalization propertie ...
京公网安备 11010802022788号
