Variable kernel density estimation - Wikipedia, the free encyclopedia
Hi everyone,
I am currently writing an algorithm using C, to perform an adaptive/variable kernel density estimation. I have been trying to use the NAG library which I am sure a lot of you are familiar with. I went through the documentation (NAG C Library - Supporting Documentation | Numerical Algorithms Group) but I was unable to find anything suitable. Does anyone have any ideas to point me in the right direction? Cheers!!
What about the link at the bottom of the wiki page?
libAGF - A library for multivariate, adaptive kernel density estimation.
If you dance barefoot on the broken glass of undefined behaviour, you've got to expect the occasional cut.
If at first you don't succeed, try writing your phone number on the exam paper.
I did actually check that out. That library is for C++, which in my understanding was quite a bit slower than C? Am I wrong in that assumption? This algorithm needs to be able to run extremely fast. Cheers!
Hardly, but even if it was, reading and adapting an existing open source library is often a good place to startOriginally Posted by statsrabbit
Pre-optimization is the root of all evil. How do you know the C++ implementation wont be fast enough? Sure, an ideal C implementation would probably be faster than an ideal C++ implementation. But you have to make that C implementation, and you will have to make it fast. You will also have lots of debugging to do. You will probably save tons of development time by using the existing C++ implementation, and it may be more than fast enough for your needs.
> That library is for C++, which in my understanding was quite a bit slower than C?
Well there's no shortage of the word "fast" if you follow the library links.
You could be done coding by the end of the week if you use that library, then you can start analysing your data.
Or you could spend weeks/months rolling your own code, debugging it, optimising it before you get to analyse any data at all. If the only difference between your very late C answer and early C++ answer is a few seconds of run-time, it's going to be a seriously long time before that pays off.
Besides, that library is GPL'ed, so you can just grab the code and start tweaking it, if it isn't quite good enough.
If you dance barefoot on the broken glass of undefined behaviour, you've got to expect the occasional cut.
If at first you don't succeed, try writing your phone number on the exam paper.
FYI:
I glanced at the source code of libagf; and, the only C++ use that jumps out is the use of pass by reference (&).
I saw no classes in the three headers I checked.
Tim S.
Wow, thanks everybody for the help! So I am a bit of a noob at both C and C++, and basically right now I am storyboarding what needs to be done step by step and finding the appropriate libraries, etc to draw my functions from. I am open to using either language, I had just been told that C was faster on the whole, and the purpose that we need to use this algorithm for require the run time to be ridiculously low..
Tim, by saying that there are no headers in that class, did you mean that I would be able to port it to C if I so wished?
Is it possible to have parts in C and parts in C++? Cheers!!
I could port it to C if the project does not use the C++ keyword "class"; no idea if you could.
Yes, it is possible to mix C with C++ code.
NOTE: I said there where NO classes in the three header files I checked; not what you stated.
Tim S.
Last edited by stahta01; 08-29-2011 at 06:54 PM. Reason: Grammer error
Oh sorry about that, Tim. Ok fantastic, I think I will stick to just coding in C++ for now. Thanks for the help!
