BriefGPT.xyz
Apr, 2011
对受正则化目标训练的模型输入添加噪音
Adding noise to the input of a model trained with a regularized objective
HTML
PDF
Salah Rifai, Xavier Glorot, Yoshua Bengio, Pascal Vincent
TL;DR
研究如何通过加入噪声对神经网络进行正则化及对其泛化性能的影响进行分析,并探讨不同类型的正则化技术对模型参数影响的先验分布。
Abstract
regularization
is a well studied problem in the context of
neural networks
. It is usually used to improve the
generalization
performance w
→