Re: solutions looking for problems to solve (on polygons) (ray shooting and visibility)

"Ralph Boland" <rpboland@xxxxxxxxx> wrote in message
news:60307a79-0ee1-4504-bc2c-67fbd20f9574@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

I wrote:
You have developed algorithms without knowing any problems
to which they apply, and you are willing to publish your results ...
and you responded:
Actually ray shooting has numerous applications in computational
geometry.
<snip>

I am aware of such applications. My comment was that you stated
specifically that you know of no problems to which *your* algorithms
apply. And you mention this again:

There are numerous papers on visibility problems on polygons and
they list numerous applications of their algorithms though I have
not seen the specific problem I solve addressed. In any case, in
my opinion, the problems my algorithms solve are of sufficiently
fundamental nature that I would have designed them even if I
knew of no papers on their subjects and no applications to which
they apply.

I review papers submitted to a few journals in the field. If I see one
that says "Here is a new algorithm, but I have no application", I would
reject it for publication immediately. At least try to come up with
some relevant application on your own that justifies the algorithm
(and a publication related to it). You mention later in this post that
you believe you can improve on some visibility-graph algorithms. At
least use that to justify your new methods. If you have studied
this type of problem for a few years, then clearly you should be able
to formulate applications to which the methods apply and show they
are superior.

You wrote:
If you have knowledge (or ideas) of applications of these algorithms,
preferably in computer graphics,
I would appreciate hearing about them.
and now write:
It's a mystery to me why you think I haven't done this.

Because your statement asks participants in this newsgroup to
identify applications? I have been participating in this newsgroup
for about 17 years now. When I see such a statement, it is invariably
someone wanting others to do the leg work for him. My apologies that
I did not read more into your statement.

I also found several papers on visibility graphs of polygons but
none come close to the performance achieved by my algorithms.

Then this is what you should stress in a publication. The emphasis
is on the improved performance *and* you need implementations
to compare. Data structures based on asymptotic analysis are fine
in theory, but in some practical applications, the benefits of the
"better" asymptotic algorithm are not noticed because the N is
not sufficiently large or the constant is extremely large (due to
whatever specific code is involved, cache misses and other
hardware issues that can affect the constant, etc.) There is also
the tradeoff of engineering time to implement a complicated
algorithm versus less time to implement a (slower) simpler one.
As an example, consider that there appears not to be a public
implementation of Chazelle's linear-time polygon triangulation
algorithm. If it takes 1 year to implement such a thing, I'd
rather spend 1 week and implement ear clipping.

Given that none of the papers on ray shooting in polygons or
visibility graphs on polygons I looked at, gave an implementation
and given that, for the 3 optimal ray shooting algorithms at least,
I have not been able to find an actual implementation anywhere,
why do you put me to the fire for failing to do so?

Because your paper would be one more that claims theoretical
superiority without a demonstration of practical superiority?

What ray shooting are you referring to? For the standard
point-in-polygon test, there are implementations to be found.
If you have studied this type of problem in detail, how difficult
would it be to implement the 3 optimal ray shooting algorithms,
and then yours, and then compare? Is this not the point of
research, of getting a PhD and postdoc?

The percentage of papers in computational geometry that come
with an implementation is very small; admittedly, computer graphics
does somewhat better. My graduate superviser has his name on
about 300 published papers but, as far but as far as I know, he
has never implemented any of the algorithms in those papers
nor had any of his graduate students do so.

Is this justification for not implementing your own? I am an engineer
who wants real solutions in a practical environment. I take issue
with the academics who avoid the practical matters. Many of us
who frequent this group are well aware of the computational
geometry algorithms that assume exact, real-valued arithmetic
(i.e. no numerical round-off errors). I have no problem with these
theoretical results, but I guarantee you that developing robust
implementations in the presence of inexact arithmetic is not
trivial and needs more research.

I have spent two years designing these algorithms and
I will spend probably another year polishing them and
writing the necessary papers. So far I have recieved
appx $30,000 for these efforts (a post-doc's salary for
a year) For the rest I will receive 0$ in compensation
and indeed have to pay out of my pocket to travel to
conferences to present papers on them (and pay
conference fees). I feel I have paid my due on these
algorithms and have drawn the limit at implementing them
for free. If it turns out that no one else feels these
algorithms are worth implementing then they will go
unimplemented.

This focus of your original post, and now here, is what I find
bothersome. My gawd, man, nobody owes you anything. I've
been through the Phd stuff and the postdoc stuff (well, as
another Phd) and I never felt like I got the bad end of the
stick. I despise the term "I have paid my due". No, you
have not--there is no such thing. Finish your postdoc, get
your publications, and get on with the next job. Get one for
which you enjoy doing the work, put a lot of effort into it,
and consider that the large salary you get is a pleasant
side effect of being given the opportunity to contribute in ways
that most people do not have the chance to do.

Allow me to answer your question this way:
You show no interest in my algorithms and
make no effort to answer my questions in any way
or point to where I might find answers.
But you have responded in a manner I interpret as
sarcastic and counterproductive.
Have I understood properly?

No, you have not.

Your post gave no specific and useful information for me to
help you.

I have implemented visibility graph algorithms in practical software.
I am actually interested in your methods. Submit your paper
to Journal of Graphics Tools (A.K. Peters publisher) and ask
that I be a reviewer.

--
Dave Eberly
http://www.geometrictools.com



.

Relevant Pages

  • Re: Havent done anything real with OOP yet.
    ... > find handbooks for specific applications, the web, My First Book on ... etc.) besides the solution algorithms. ... one can write a subsystem for GUI ... > OOPL and I think Java books get their readers to flashy graphical stuff as ...
    (comp.object)
  • Re: IL Code Security
    ... If your boss is so convinced that the algorithms are more secure in C++ ... > more critical applications to be easily read via a disassembler (even ... > 1) Obfuscation: Although this makes it more difficult to read, ... > assemblies wouldn't execute after being obfuscated. ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: emergent/swarm/evolutionary systems etc
    ... > all of the applications being developed and used are far 'stronger' than ... > considerably 'stronger' than I would desire to use. ... networks), genetic algorithms, naive bayesian statistics, etc., the ... Don't think of Python as a scripting language. ...
    (comp.lang.python)
  • Call for Chapters: Variants of Evolutionary Algorithms for Real-World Applications
    ... Variants of Evolutionary Algorithms for Real-World Applications ... Started as a mere academic curiosity, Evolutionary Algorithms (EAs) ... Algorithms, Evolutionary Programming, Evolution Strategies and Genetic ...
    (comp.ai.alife)
  • Call for Chapters: Variants of Evolutionary Algorithms for Real-World Applications
    ... Variants of Evolutionary Algorithms for Real-World Applications ... Started as a mere academic curiosity, Evolutionary Algorithms (EAs) ... Algorithms, Evolutionary Programming, Evolution Strategies and Genetic ...
    (comp.ai.genetic)
Loading