人工智能培训

搜索

深度学习论文:线性回归的改进PAC-贝叶斯界(Improved PAC-Bayesian Bounds for Linear Regression)

[复制链接]
deshuo 发表于 2019-12-9 14:01:02 | 显示全部楼层 |阅读模式
deshuo 2019-12-9 14:01:02 126 0 显示全部楼层
深度学习论文:线性回归的改进PAC-贝叶斯界(Improved PAC-Bayesian Bounds for Linear Regression)在本文中,我们改进了Germain等人的线性回归的PAC-贝叶斯误差界。 [10]。改进是双重的。首先,建议的误差范围更严格,并且在选择好的温度参数下收敛到泛化损失。其次,误差范围还包含训练数据,这些数据不是独立采样的。尤其是,误差范围适用于由众所周知的动力学模型类别(例如ARX模型)生成的某些时间序列。
In this paper, we improve the PAC-Bayesian error bound for linear regressionderived in Germain et al.[10].The improvements are twofold.First, theproposed error bound is tighter, and converges to the generalization loss witha well-chosen temperature parameter.Second, the error bound also holds fortraining data that are not independently sampled.In particular, the errorbound applies to certain time series generated by well-known classes ofdynamical models, such as ARX models.深度学习论文:线性回归的改进PAC-贝叶斯界(Improved PAC-Bayesian Bounds for Linear Regression)
URL地址:https://arxiv.org/abs/1912.03036     ----pdf下载地址:https://arxiv.org/pdf/1912.03036    ----深度学习论文:线性回归的改进PAC-贝叶斯界(Improved PAC-Bayesian Bounds for Linear Regression)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

deshuo当前离线
新手上路

查看:126 | 回复:0

快速回复 返回顶部 返回列表