arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2007.05434cs.LGmath.STstat.MLstat.TH
收藏

宽神经网络的Monte Carlo丢弃特性

Characteristics of Monte Carlo Dropout in Wide Neural Networks

Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS(弗劳恩霍夫智能分析与信息系统研究所);Fraunhofer Center for Machine Learning(弗劳恩霍夫机器学习中心);Ruhr-University Bochum(波鸿鲁尔大学)

Joachim Sicking, Maram Akila, Tim Wirtz, Sebastian Houben, Asja Fischer

摘要:

蒙特卡罗(MC)丢弃是神经网络(NN)不确定性估计的一种最新方法。它被解释为近似执行贝叶斯推理。基于以前的工作的高斯过程的近似宽和深神经网络随机权重,我们研究了广泛的未经训练的神经网络下dropout的极限分布更严格,并证明了他们也收敛到高斯过程的固定集的权重和偏见。我们勾勒出一个论点,即这个属性也可能适用于无限宽的前馈网络,这些网络是用(全批)梯度下降训练的。该理论是对比的实证分析中,我们发现相关性和非高斯行为的预激活有限宽度NN。因此,我们研究(强)相关的预激活如何在具有强相关权重的NN中诱导非高斯行为。

英文摘要:

Monte Carlo (MC) dropout is one of the state-of-the-art approaches for uncertainty estimation in neural networks (NNs). It has been interpreted as approximately performing Bayesian inference. Based on previous work on the approximation of Gaussian processes by wide and deep neural networks with random weights, we study the limiting distribution of wide untrained NNs under dropout more rigorously and prove that they as well converge to Gaussian processes for fixed sets of weights and biases. We sketch an argument that this property might also hold for infinitely wide feed-forward networks that are trained with (full-batch) gradient descent. The theory is contrasted by an empirical analysis in which we find correlations and non-Gaussian behaviour for the pre-activations of finite width NNs. We therefore investigate how (strongly) correlated pre-activations can induce non-Gaussian behavior in NNs with strongly correlated weights.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑