Sampling Hidden Parameters from Oracle Distribution

Sampling Hidden Parameters from Oracle Distribution
复制标题

从 Oracle 分布中采样隐藏参数

DOI：
10.1007/978-3-319-11179-7_68
复制
发表时间：
2014
期刊：
--
影响因子：
--
通讯作者：
Noboru Murata
Noboru Murata
中科院分区：
--
文献类型：
--
作者：
Sho Sonoda;Noboru Murata

文献摘要

被引文献

相似文献

提出了一种新的神经网络采样学习方法。从神经网络的积分表示导出，引入了隐藏参数的概率分布。一般来说，从预言分布中严格采样存在数值困难，因此还开发了线性时间采样算法。数值实验表明，当通过预言分布初始化隐藏参数时，与通过正态分布初始化参数时相比，反向传播更快地收敛到更好的参数。
A new sampling learning method for neural networks is proposed. Derived from an integral representation of neural networks, anoracleprobability distribution of hidden parameters is introduced. In general rigorous sampling from the oracle distribution holds numerical difficulty, a linear-time sampling algorithm is also developed. Numerical experiments showed that when hidden parameters were initialized by the oracle distribution, following backpropagation converged faster to better parameters than when parameters were initialized by a normal distribution.