一个带有 Dropout 界限的 PAC-Bayesian 教程

Jul, 2013

A PAC-Bayesian Tutorial with A Dropout Bound

David McAllester

TL;DR这篇论文简要介绍了现有的PAC-Bayesian理论，重点关注三种泛化界限及其应用，可以有效地处理规则参数及提供训练保障。

Abstract

This tutorial gives a concise overview of existing pac-bayesian theory focusing on three generalization bounds. The first is an occam bound