Erroneous documentation for auto_basis

[tatephughes]

In the documentation for auto_basis, it says

If the manifold is a plane, and regular > 0, then basis functions are placed regularly within the bounding box of data, with the smallest number of basis functions in each row or column equal to the value of regular in the coarsest resolution (note, this is just the smallest number of basis functions). Subsequent resolutions have twice the number of basis functions in each row or column.

However, the numbers here seem to be incorrect. The part of the code (lines 460 and 461) for the number of basis functions in placed in resolution i is round(nx*(3^(i)), where i ranges from 1 to nres, (assuming x is the wider dimension, so nx<-regular on line 416, the same applies if y is the wider dimension of course).

This means that the number of basis function at the coarsest resolution is actually 3 times the value of regular, and subsequent resolutions have thrice the number of basis functions in each row/column.

Really sorry if I have misunderstood the documentation of course!

[andrewzm]

Thanks for pointing this out, this will be fixed in the next release. Best, Andrew

…

On Thu, Sep 19, 2024 at 8:51 PM Evan Tate Paterson Hughes < ***@***.***> wrote: In the documentation for auto_basis <https://www.rdocumentation.org/packages/FRK/versions/2.3.1>, it says If the manifold is a plane, and regular > 0, then basis functions are placed regularly within the bounding box of data, with the smallest number of basis functions in each row or column equal to the value of regular in the coarsest resolution (note, this is just the smallest number of basis functions). Subsequent resolutions have twice the number of basis functions in each row or column. However, the numbers here seem to be incorrect. The part of the code (lines 460 and 461) for the number of basis functions in placed in resolution i is round(nx*(3^(i)), where i ranges from 1 to nres, (assuming x is the wider dimension, so nx<-regular on line 416, the same applies if y is the wider dimension of course). This means that the number of basis function at the coarsest resolution is actually 3 times the value of regular, and subsequent resolutions have *thrice* the number of basis functions in each row/column. Really sorry if I have misunderstood the documentation of course! — Reply to this email directly, view it on GitHub <#9>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABEU3PQTLQAJTFSLDLQ623ZXKUDLAVCNFSM6AAAAABOPUK3T6VHI2DSMVQWIX3LMV43ASLTON2WKOZSGUZTKOJYG4ZDCMA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>