Guassian Process Regression (fitrgp()): Setting KernelScale as a constant without specifying FitMethod as 'none'

5 次查看（过去 30 天）

显示更早的评论

Sterling Baird 2021-4-15

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/802936-guassian-process-regression-fitrgp-setting-kernelscale-as-a-constant-without-specifying-fitmeth

评论： Sterling Baird 2021-5-8

I'd like to set KernelScale as constant similar to how ConstantSigma can be specified; however, if I choose FitMethod='none', then Beta and other variables aren't analytically solved for. Assuming this isn't a theoretically unsound request, how do I set KernelParameters ( KernelScale and Sigma) as fixed?

Related: How to check the kernel parameter values of a Gaussian process regression (GPR) model after training

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

回答（1 个）

伟张 2021-5-8

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/802936-guassian-process-regression-fitrgp-setting-kernelscale-as-a-constant-without-specifying-fitmeth#answer_694890

If you have one more input, the software set a fixed kernalscale by default.Thus you don"t have to worry about this problem.

2 个评论
显示无隐藏无

伟张 2021-5-8

of course it is also up to what kernel function you choose

Sterling Baird 2021-5-8

I think you misunderstand. I'd like to specify KernelScale and Sigma myself, not go with the default. But I want it to still solve for Beta, etc.

请先登录，再进行评论。

请先登录，再回答此问题。

类别

AI and Statistics Statistics and Machine Learning Toolbox Regression Gaussian Process Regression

在 Help Center 和 File Exchange 中查找有关 Gaussian Process Regression 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Translated by