Re: Information-Theory functions.

On Sep 28, 2006, at 7:55 PM, Muthiah Annamalai wrote:

> Hi there,
> I have my first cut functions on basic functions in
> Information Theory; the package is a nice candidate
> for inclusion into Octave-forge. If Paul or David
> can approve it, I can gohead and fix it according
> to Octave-forge specifications.

1) If you structure the initial comments as follows
then it won't show up in the help string.

## Copyright (C) ...
##
## This program is free software ...
## ...

## normal function help
## ...

2) If you put tests of the following form after your
function they will be automatically run on install:

%!assert(expr, expected, tol)

See 'help test' for details.

3) You will need to use the new package directory
structure and metadata files.


4) texinfo is the usual format for documentation
(and someday even I might start using it 8-)

5) Your 'see also' help needs to be in the same
comment block as the rest of the help.

6) Are you sure the info-theory.m file needs to
have copies of all your functions?  Also, it
probably shouldn't have an extension since it
is available on the octave command line.

7) In mutualinformation, rather than:

   idx_rng = 1:cols
   tVal = XY(i,idx_rng)./(X(idx_rng)*Y(i))

You can write:

   tVal = XY(i,:)./(X*Y(i))

Also, it seems to have one too many end statements.

8) Rather than size(X)(2) you can write size(X,2)


I haven't checked whether these functions duplicate
existing matlab functions.

- Paul