It would be good to explicitly compare the work here with the dual package, specificially the nice example in the README file which shows how the derivative is vectorized, unlike the approach used here. I don't think it's better or worse, just different. But in any case it would be good to document this clearly. For the record, Satore uses the following example:
library("dual")
x <- dual(f = 1.5, grad = c(1:0, 0))
y <- dual(f = 0.5, grad = c(0:1, 0))
z <- dual(f = 1.0, grad = c(0, 0:1))
exp(z - x) * sin(x)^y / x
#> Real: 0.403847
#> Duals: -0.658759 -0.001013 0.403847
See how the package nicely gives a (vector) gradient at $(1.5,0.1,1)$. It would be nice to reproduce this with clifford idiom, if only to show how it might be done [and also that my approach currently needs to use %/%, see issue #96].
RobinHankin
commented
3 months ago
It would be good to explicitly compare the work here with the
dualpackage, specificially the nice example in theREADMEfile which shows how the derivative is vectorized, unlike the approach used here. I don't think it's better or worse, just different. But in any case it would be good to document this clearly. For the record, Satore uses the following example:See how the package nicely gives a (vector) gradient at $(1.5,0.1,1)$. It would be nice to reproduce this with
cliffordidiom, if only to show how it might be done [and also that my approach currently needs to use%/%, see issue #96].