Re: too many files!
Joel Pitt
> Speaking of which - I was toying with the notion of adding a kernel form to
> r.mdig.kernel - is it as simple as adding a cdf to kernel_math.c? Ideally
> it would be nice to have a mixed model, which is just a summation of a
> negative exponential and an inverse power curve - it takes the form N = a *
> exp(b*dist) + c * dist^-d. I have to think about that a bit, but I'd like
> to try and stick it in there. That form provided a significantly better fit
> than other available options explored thusfar. I'll try it with a different
> one for now though.
That should be possible, although you'd have to work out how to
normalise the mixed model to keep results consistent. You might be
able to get away with not doing this, but it may make interpreting
results difficult. If you can come up with a CDF, or a numerical
algorithm that acts like a CDF (see the Clark method in kernel_math.c)
then it should be easy enough to add. You'll also have to add the
kernel to the option parsing code in main.c
I can help with this stuff, but just be warned that C is another level
pedantic coding that is often more tedious and frustrating than
Python.
> BTW - Richard and I have been invited over to melbourne to discuss with some
> people the work we've been doing, and they might be interested in employing
> popMod (at a simpler level) to create scenarios with which to test detection
> strategies. Do you have any plans as far as further development ideas that
> I might want to mention to them? Specifically, probably anything that would
> make the user interface easier for less technically-oriented people would
> probably be of interest, which might be relevant to the stuff you are
> planning to do for Sue. Just thought I'd mention it.
There will be a web interface (which can be run and accessed locally,
i.e. no server system needed), and a potentially a system to run MDiG
across multiple computers to speed up processing. The web interface
will initially just let you submit models and browse results, but it
should also allow one to build models to avoid having to edit XML
directly.
There is also a Windows version that I'll be working on and this
should be released by the end of June (hopefully sooner).
Cheers,
Joel