Hi all,
I'm trying to get a simple, linear decision surface from e1071's svm.
I've run it like this:
svm(as.factor(slow) ~ SLICE.3 + PSGR.7 + SOLUTIONS.6 + DR.10, y,
kernel='linear', cost=1e6, class.weights=c('FALSE'=1, 'TRUE'=10))
According to the docs, kernel='linear' has a kernel u'v. Since I have 4
independent variables, I'd expect to have four coefficients plus a
threshold, with 4 total degrees of freedom. But the only numeric
vectors of length 4 in the result are the scaling and center, and those
are done before the fitting so each one has zero mean and unit variance.
I know svms don't need to put every point through the kernel function,
and can even handle infinite dimensional kernels. But don't they need
to compute the coefficients?
Best,
Martin
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
