BriefGPT.xyz
Dec, 2013
利用概率最大化单元优化深度神经网络
Improving Deep Neural Networks with Probabilistic Maxout Units
HTML
PDF
Jost Tobias Springenberg, Martin Riedmiller
TL;DR
提出了一种概率版本的maxout单元，该单元采用一组线性变换的池化操作，具有可以改进的不变性特性，在三项具有挑战性的图像分类基准测试中，表现优异，证明了本文提出的probout单元是有效的。
Abstract
We present a probabilistic variant of the recently introduced maxout unit. The success of
deep neural networks
utilizing maxout can partly be attributed to favorable performance under
dropout
, when compared to re
→