All geek questions in one place

R carriage stop criterion

I use the caret package with the nnet method in the classification problem, and I would like to know what is the stopping criterion used in nnet to stop training in order to avoid over-matching.

I did some research and I found that the RSNNS package does not have a stopping criterion, but it is considered for future work, see here: https://www.jstatsoft.org/index.php/jss/article/view/v046i07/ v46i07.pdf

But for the nnet package, I did not find any information about the stopping criteria used or if it uses it. So my question. What is the stopping criterion used in the nnet package?

Thanks!!!

+5
source share
1 answer

The stopping criteria used in the nnet package is determined by the abstol and reltol the nnet() function.

 abstol Stop if the fit criterion falls below abstol, indicating an essentially perfect fit. reltol Stop if the optimizer is unable to reduce the fit criterion by a factor of at least 1 - reltol.

(copied from https://cran.r-project.org/web/packages/nnet/nnet.pdf )

Their default values ​​are: abstol = 1.0e-4; reltol = 1.0e-8

The correspondence criterion is some error between the truth of the earth and the output of the network. This is part of the objective function for iterative minimization.

A question arises regarding the compliance criterion with a partial answer: https://stats.stackexchange.com/questions/156117/what-is-the-value-of-fitting-criterion-on-the-nnet-package-in-r

+2
source

Source: https://habr.com/ru/post/1243343/

More articles:


All Articles