sabbatical project

[thomas.l.scofield]

Hello.

I haven't contributed that much to Octave as of yet.  My main contribution was to alter imread from its state in octave-forge so that it would use the GraphicsMagick library, and to significantly re-write imwrite so that it would use that same library, and so it would become much more Matlab-like.

While these things were fun, and helped me understand the Octave culture and structure to some extent, I'm really more interested in implementing mathematical algorithms and/or developing packages to do state-of-the-art calculations for mathematical problems.  As a college professor with enough years of service, I am eligible for a sabbatical, approximately half of a school year (plus summers on either end of the year) to work on some project I propose and deemed worthy by my school.  I am seeking both advice and help.

When it comes to advice, I would appreciate help with choosing a package to work on.  I have done some work in imaging and am interested in doing more, and in that vein I have thought it would be interesting to work on packages of the following types: general image processing (contributions to the image package), a package for compressive sampling, a package for level set methods, and a wavelet package; I believe the latter three would be new starts in Octave.  Having identified these possible packages, I'm afraid the standard "work on what you are interested in" will not help me here.  What would help is knowing whether packages like these are getting some attention already, whether there would be much demand, and insight into the dependency structure (i.e., I do not yet know enough about level set methods to know if it would helpful to have a wavelet package already in place).

As for help, at my school a sabbatical proposal must be accompanied by "a letter of endorsement from a scholar outside the college (i.e., a non-Calvin scholar or colleague who knows the relevant field of scholarship, is familiar with the applicant's previous work, and is in a position to assess both the importance of the proposed project and the likelihood of its being brought to successful completion)."  I know a post at octave-forge would probably be better, since any work on a package I do is likely to end up there, but I think only people in the octave-maintainers list are going to be at all familiar with the scant work I've done.  Do any of you feel qualified to write such a letter?

I'm afraid I'm putting this together rather late.  I have to have all materials for the proposal submitted to college officials by Friday Sept. 20 (1 week from today).

Thomas L. Scofield

--------------------------------------------------------

Associate Professor

Department of Mathematics and Statistics

Calvin College

--------------------------------------------------------

[dbateman]

thomas.l.scofield wrote:

Hello.

I haven't contributed that much to Octave as of yet.  My main  
contribution was to alter imread from its state in octave-forge so  
that it would use the GraphicsMagick library, and to significantly re-
write imwrite so that it would use that same library, and so it would  
become much more Matlab-like.

While these things were fun, and helped me understand the Octave  
culture and structure to some extent, I'm really more interested in  
implementing mathematical algorithms and/or developing packages to do  
state-of-the-art calculations for mathematical problems.  As a  
college professor with enough years of service, I am eligible for a  
sabbatical, approximately half of a school year (plus summers on  
either end of the year) to work on some project I propose and deemed  
worthy by my school.  I am seeking both advice and help.

When it comes to advice, I would appreciate help with choosing a  
package to work on.  I have done some work in imaging and am  
interested in doing more, and in that vein I have thought it would be  
interesting to work on packages of the following types: general image  
processing (contributions to the image package), a package for  
compressive sampling, a package for level set methods, and a wavelet  
package; I believe the latter three would be new starts in Octave.  
Having identified these possible packages, I'm afraid the standard  
"work on what you are interested in" will not help me here.  What  
would help is knowing whether packages like these are getting some  
attention already, whether there would be much demand, and insight  
into the dependency structure (i.e., I do not yet know enough about  
level set methods to know if it would helpful to have a wavelet  
package already in place).

As for help, at my school a sabbatical proposal must be accompanied  
by "a letter of endorsement from a scholar outside the college (i.e.,  
a non-Calvin scholar or colleague who knows the relevant field of  
scholarship, is familiar with the applicant's previous work, and is  
in a position to assess both the importance of the proposed project  
and the likelihood of its being brought to successful completion)."  
I know a post at octave-forge would probably be better, since any  
work on a package I do is likely to end up there, but I think only  
people in the octave-maintainers list are going to be at all familiar  
with the scant work I've done.  Do any of you feel qualified to write  
such a letter?

I'm afraid I'm putting this together rather late.  I have to have all  
materials for the proposal submitted to college officials by Friday  
Sept. 20 (1 week from today).

Thomas L. Scofield
--------------------------------------------------------
Associate Professor
Department of Mathematics and Statistics
Calvin College
--------------------------------------------------------

Some of my thoughts and Jaroslav's about what the goals for 3.3 should be are in the thread

http://www.nabble.com/forum/ViewPost.jtp?post=19182313

Though Michael has also started in on classdef support, so I suppose that should also be considered a 3.3 goal. The only really mathematical subject there is the non-linear minimizers, though I think Jaroslav has started on that. For the less mathematical tasks I still believe the tree walker evaluator class is the next big thing that is needed to get better acceptance of Octave as its a necessary step towards a profiler, a compiler and a JIT.

Octave supports matlab mex files and so wavelab works fine under Octave, so its a little difficult to say that Octave needs a wavelet toolbox. As for other things, well I'd look for inspiration in missing toolbox or core functions. Check

 http://wiki.octave.org/wiki.pl?MissingMatlabFunctions

for the list of missing Matlab 2008a core function in Octave 3.0.1 with their current state with respect to 3,1.51+

Regards
David

[Søren Hauberg]

fre, 12 09 2008 kl. 16:31 -0400, skrev Thomas L. Scofield:

> When it comes to advice, I would appreciate help with choosing a
> package to work on.  I have done some work in imaging and am
> interested in doing more, and in that vein I have thought it would be
> interesting to work on packages of the following types: general image
> processing (contributions to the image package), a package for
> compressive sampling, a package for level set methods, and a wavelet
> package; I believe the latter three would be new starts in Octave.
> Having identified these possible packages, I'm afraid the standard
> "work on what you are interested in" will not help me here.  What
> would help is knowing whether packages like these are getting some
> attention already, whether there would be much demand, and insight
> into the dependency structure (i.e., I do not yet know enough about
> level set methods to know if it would helpful to have a wavelet
> package already in place).

For the last couple of months I've been wanting to work on functions for
level set segmentation, but I haven't had the time. Specifically, I'm
very interested in the framework for shape priors in level set
segmentation developed by Daniel Cremers and colleagues. So, if this is
something you would like to spend time on, I'd be very interested to
help as much as time allows for.

> As for help, at my school a sabbatical proposal must be accompanied by
> "a letter of endorsement from a scholar outside the college (i.e., a
> non-Calvin scholar or colleague who knows the relevant field of
> scholarship, is familiar with the applicant's previous work, and is in
> a position to assess both the importance of the proposed project and
> the likelihood of its being brought to successful completion)."

I only have a masters degree in Computer Science, so I don't think I'm
qualified to help you with this part (hopefully, I'll start on a phd in
a month or two). But if I can help you here anyway, do let me know.

Søren

[Jaroslav Hajek-2]

On Sat, Sep 13, 2008 at 2:02 PM, dbateman <dbateman@...> wrote:

>
>
>
> thomas.l.scofield wrote:
>>
>>
>> Hello.
>>
>> I haven't contributed that much to Octave as of yet.  My main
>> contribution was to alter imread from its state in octave-forge so
>> that it would use the GraphicsMagick library, and to significantly re-
>> write imwrite so that it would use that same library, and so it would
>> become much more Matlab-like.
>>
>> While these things were fun, and helped me understand the Octave
>> culture and structure to some extent, I'm really more interested in
>> implementing mathematical algorithms and/or developing packages to do
>> state-of-the-art calculations for mathematical problems.  As a
>> college professor with enough years of service, I am eligible for a
>> sabbatical, approximately half of a school year (plus summers on
>> either end of the year) to work on some project I propose and deemed
>> worthy by my school.  I am seeking both advice and help.
>>
>> When it comes to advice, I would appreciate help with choosing a
>> package to work on.  I have done some work in imaging and am
>> interested in doing more, and in that vein I have thought it would be
>> interesting to work on packages of the following types: general image
>> processing (contributions to the image package), a package for
>> compressive sampling, a package for level set methods, and a wavelet
>> package; I believe the latter three would be new starts in Octave.
>> Having identified these possible packages, I'm afraid the standard
>> "work on what you are interested in" will not help me here.  What
>> would help is knowing whether packages like these are getting some
>> attention already, whether there would be much demand, and insight
>> into the dependency structure (i.e., I do not yet know enough about
>> level set methods to know if it would helpful to have a wavelet
>> package already in place).
>>
>> As for help, at my school a sabbatical proposal must be accompanied
>> by "a letter of endorsement from a scholar outside the college (i.e.,
>> a non-Calvin scholar or colleague who knows the relevant field of
>> scholarship, is familiar with the applicant's previous work, and is
>> in a position to assess both the importance of the proposed project
>> and the likelihood of its being brought to successful completion)."
>> I know a post at octave-forge would probably be better, since any
>> work on a package I do is likely to end up there, but I think only
>> people in the octave-maintainers list are going to be at all familiar
>> with the scant work I've done.  Do any of you feel qualified to write
>> such a letter?
>>
>> I'm afraid I'm putting this together rather late.  I have to have all
>> materials for the proposal submitted to college officials by Friday
>> Sept. 20 (1 week from today).
>>
>> Thomas L. Scofield
>> --------------------------------------------------------
>> Associate Professor
>> Department of Mathematics and Statistics
>> Calvin College
>> --------------------------------------------------------
>>
>>
>>
>
> Some of my thoughts and Jaroslav's about what the goals for 3.3 should be
> are in the thread
>
> http://www.nabble.com/forum/ViewPost.jtp?post=19182313
>
> Though Michael has also started in on classdef support, so I suppose that
> should also be considered a 3.3 goal. The only really mathematical subject
> there is the non-linear minimizers, though I think Jaroslav has started on
> that.

So far I have a working fzero and buggy fsolve (a complete m-file
implementation). I have decided to postpone this somewhat, because it
is an optimization-related project and thus related to our research
grant and thus I can devote some of my paid time to that.
Unfortunately, right now I have higher priority stuff to do at work.
I wouldn't mind collaboration; and even after I finish the
implementation there will be a lot of work left because I want to
contribute some intensive testbeds for optimization funcs (fsolve, sqp
etc) into Octave-Forge, probably into the benchmarks package.


> For the less mathematical tasks I still believe the tree walker
> evaluator class

This may not be the best place, but can you please clue me what a tree
walker should do?

> is the next big thing that is needed to get better
> acceptance of Octave as its a necessary step towards a profiler, a compiler
> and a JIT.
>
> Octave supports matlab mex files and so wavelab works fine under Octave, so
> its a little difficult to say that Octave needs a wavelet toolbox. As for
> other things, well I'd look for inspiration in missing toolbox or core
> functions. Check
>
>  http://wiki.octave.org/wiki.pl?MissingMatlabFunctions
>
> for the list of missing Matlab 2008a core function in Octave 3.0.1 with
> their current state with respect to 3,1.51+
>
> Regards
> David
> --
> View this message in context: http://www.nabble.com/sabbatical-project-tp19463651p19469898.html
> Sent from the Octave - Maintainers mailing list archive at Nabble.com.
>
>

--
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz

[dbateman3]

Jaroslav Hajek wrote:
>> For the less mathematical tasks I still believe the tree walker
>> evaluator class
>
> This may not be the best place, but can you please clue me what a tree
> walker should do?


The way the visitor methods of the tree_walker class are evaluated,
there can only be a single method that is evaluated. It would be good if
the evaluation process of the tree might be rewritten such that the
visitor methods might be overloaded dynamically. In particular the
profiler might be written as a instrumented wrapper to the existing
visitor methods in this manner.

There was a discussion between John and Muthu about this about a year or
so ago on the lists..

Cheers
David





>
>
>
>