On Sun, 14 Jul 2019, Sven Schreiber wrote:
Hi,
my discovery tour of new gretl functions goes on - thanks for adding the
conv2d function!
However, I think the documentation needs to be adapted to the new
situation.
For example, in section 18.3 of the guide (Neat tricks) the convolution
is described. It still says there that "There's no dedicated function to
do that". However, now conv2d(p,q) now does that (as a special 1d-case I
guess) and produces exactly the needed outcome.
Well spotted. The method using filter() works fine, as stated in the
Guide, but conv2d() does the job with less bother so that
cheat-sheet entry should be updated.
The function reference for conv2d itself could be a little more
explicit. For example it could mention exactly this special case of
vectors as inputs and the relationship with polynomial multiplication.
OK.
(About 2d-convolution I'm not sure if perhaps there exist
slightly
different definitions or conventions? If so, a precise reference would
also help.)
I don't _think_ there's any ambiguity. This is mostly discussed in
the signal processing literature but I'm not aware of any canonical
reference. Our implementation hews pretty closely to matlab.
Allin