About ctmm.web

Fiorella Vilela

unread,

Jun 15, 2021, 6:29:15 PM6/15/21

to ctmm R user group

Hi everyone, I'm trying to do a kernel AKDE, my database has more than 2000 rows so to optimizing time I used the function "par_hrange_each" with 8 cores and weights = TRUE, but It never finishes to run. Do you have any idea of the problem?

Thank you.

Fiorella

Christen Fleming

unread,

Jun 16, 2021, 4:47:42 PM6/16/21

to ctmm R user group

Hi Fiorella,

When you look in the R console, do you see any updates regarding the weight optimization?

I have some notes in help('bandwidth') detailing what can slow this calculation down and what arguments/adjustments can help this:

https://ctmm-initiative.github.io/ctmm/reference/bandwidth.html#details

The extra arguments aren't yet supported by ctmmweb, but if you need some of these arguments in ctmmweb, then we can raise an issue on GitHub:

https://github.com/ctmm-initiative/ctmmweb/issues

At minimum, it should be trivial to add an '...' argument to par_hrange_each() that is passed to akde(). But first we need to determine what's slowing down your weight optimization calculation.

Best,

Chris

Fiorella Vilela

unread,

Jun 19, 2021, 1:53:17 PM6/19/21

to ctmm R user group

Hi Christen, thank you. I guees the problem is the irregularity of my database. As I told you before, I'm working only with foraging points. I tried with trace = TRUE with akde but it didn't work, it started to work and then it closed (I'm working in Rcloud). This is the one message in the console.
Default grid size of 1 minute chosen for bandwidth(...,fast=TRUE)

Christen Fleming

unread,

Jun 19, 2021, 4:02:07 PM6/19/21

to ctmm R user group

Hi Fiorella,

The computational cost of that calculation with fast=TRUE would be as if you had 1-minute sampling for the entire duration. But if you had mostly 1-hour sampling, then setting akde()'s dt argument to 1 hour (or 30 min or 15 min) would result in a large speed up with very little difference in the estimate. With only ~2,000 rows, you might even consider fast=FALSE and PC="direct", which would be slow but wouldn't take forever.

Best,

Chris

Fiorella Vilela

unread,

Jun 19, 2021, 8:15:07 PM6/19/21

to ctmm R user group

Hi Christen! It's working now! thank you!!

El miércoles, 16 de junio de 2021 a la(s) 15:47:42 UTC-5, chris.h...@gmail.com escribió:

Reply all

Reply to author

Forward