Mon 01 Dec 2014 08:43:08 PM UTC, comment #15:
Hi,
Thanks Doug, I pushed the patch after a few edits (added the bug n# and a minimal change log):
http://hg.savannah.gnu.org/hgweb/octave/rev/cc7931e8953d
Pantxo
|
Sun 30 Nov 2014 02:57:44 AM UTC, comment #14:
some of the testes were still failing so I added eps to them
Patch is attached.
(file #32573)
|
Mon 06 Oct 2014 04:47:22 PM UTC, comment #13:
Hi,
I have pushed the changeset at http://hg.savannah.gnu.org/hgweb/octave/rev/4122df73e374
Thanks.
|
Mon 06 Oct 2014 04:01:08 PM UTC, comment #12:
Ok
Here it is.
(file #32232)
|
Mon 06 Oct 2014 01:50:24 PM UTC, comment #11:
Doug-- Yes, please do.
|
Sun 05 Oct 2014 04:44:27 PM UTC, comment #10:
I compiled octave and ran a test
doug@speedy:~/octavec/octave$ hg tip
changeset: 19220:6443693a176f
bookmark: @
tag: tip
lscov failed with:
error: ASSERT errors for: assert (b2,[b, 2 * b])
Location | Observed | Expected | Reason
(1,1) 0.1203 0.1203 Abs err 1.3878e-17 exceeds tol 0
(1,2) 0.24061 0.24061 Abs err 2.7756e-17 exceeds tol 0
if I change to:
assert(b2, [b 2*b] , eps)
then It passes ok.
Do you want me to make a hg change set for this?
Doug Stewart
|
Mon 08 Sep 2014 06:33:57 AM UTC, comment #9:
I have moves the script to scripts/statistics. Closing bug report.
|
Sun 07 Sep 2014 02:45:30 PM UTC, comment #8:
OK
|
Sun 07 Sep 2014 01:22:50 PM UTC, comment #7:
I'll move the function to scripts/statistics/base.
I removed "regress" from the seealso section as it currently belongs to the statistics package, not octave core.
|
Sun 07 Sep 2014 11:56:11 AM UTC, comment #6:
Couple minor suggestions:
1) Consider putting lscov in statistics (like ols and gls) instead of optimization
2) Add regress under seealso
Otherwise, looks great. You can close the bug report.
|
Sun 07 Sep 2014 09:38:58 AM UTC, comment #5:
Hi Nir,
I have added lscov to octave build tree (http://hg.savannah.gnu.org/hgweb/octave/rev/c9113e28fae8).
If you nothing wrong in the commit I'll close this bug report
|
Sat 06 Sep 2014 05:41:02 PM UTC, comment #4:
Based on what lsqnonneg and so on have in the manual, looks like we can pretty much just copy the help text over to that
|
Sat 06 Sep 2014 05:23:25 PM UTC, comment #3:
Hi Nir,
Can you suggest a small text to introduce this function in octave manual (in chapter 25.4 maybe [1])?
[1] https://www.gnu.org/software/octave/doc/interpreter/Linear-Least-Squares.html#Linear-Least-Squares
|
Fri 05 Sep 2014 12:52:35 PM UTC, comment #2:
Your changes are generally fine. I made a few small modifications (mostly spelling and style) -- see the attached.
(file #32029)
|
Fri 05 Sep 2014 09:05:12 AM UTC, comment #1:
Hi Nir,
I have no expertise in this kind of algorithm but I can make a few formal proposals. I have attached a modified version of your function with the following mods:
- add space between function names and parens to visually differentiate function calls and matrix indexing,
- reduce indentation to 2 spaces,
- enclose boolean expressions after "if" statements between parens,
- add "endfunction" before tests,
- add a few input checkings so that calling the function e.g. "lscov (1)" returns a useful error message,
Finally I also reorganized the doc so that it looks more like other octave functions help:
- list the different ways to call the function in the header, so that it is obvious what input parameter is optional or not,
- remove the input/output sections in favor of dedicated paragraphs. As an example of a function that have a similar interface, see the help string for "fsolve" function.
Can you tell me if you agree with my proposal?
(file #32028)
|
Sun 31 Aug 2014 11:43:47 PM UTC, original submission:
Candidate implementation attached
|