Numerical integration - optimisation of code

AI Thread Summary
The discussion focuses on optimizing numerical integration for a non-analytic function resembling a Gaussian, aiming to reduce computation time from five seconds per curve to real-time performance. Suggestions include avoiding regular grids and utilizing Gaussian quadrature for more efficient area calculations, particularly in regions of high slope. The importance of leveraging MATLAB's built-in numerical quadrature routines is emphasized, as they provide adaptive solutions that can handle complex integration tasks. Additionally, the potential for using symmetry in the function to simplify calculations is noted, although the user is constrained to MATLAB for integration with existing code. Overall, the thread highlights the need for efficient computational techniques in numerical integration.
mikeph
Messages
1,229
Reaction score
18
Hello

I have a function which is very similar in shape to a Gaussian, except it is not a distribution and it is not analytic, so I can at best calculate a single point on the curve at a time. (In general it is a convolution of different distributions but this is not important).

I need to find the area under this curve accurately, using the least amount of calculations (I need to find the area under ~1,000 curves and right now it is taking about 5 seconds per curve to get within about 1% accuracy, ideally the code needs to perform this in real time! ).My first try was to use a regular grid, eg. -10:0.01:10 for 2001 points, then the area (under)approximated by the sum of the values/100, the resolution. Varying the range and resolution of integration get me close to the "true" value but it takes too long, and I think I am wasting high resolution on the tail.
My next idea for optimisation is to have a grid which has more resolution at the areas of highest slope (does this make mathematical sense? I think...). I am having trouble justifying this and do not really know where to begin to implement it.

If anyone could point me to the right direction, I am sure this has been done before.

Regards,
Mike
 
Mathematics news on Phys.org
Some suggestions:
  • Don't use a regular grid. Look into something like Gaussian quadrature.
  • Is the function symmetric? If so, find the center and integrate half of the domain.
  • That you said "-10:0.01:10" suggests you are doing this in Matlab. Matlab is 10-100 times slower than C or Fortran.
 
Added clarification to 2nd bullet in case it wasn't obvious to the OP.
D H said:
Some suggestions:
  • Don't use a regular grid. Look into something like Gaussian quadrature.
  • Is the function symmetric? If so, find the center and integrate half of the domain, and then double the result[/color].
  • That you said "-10:0.01:10" suggests you are doing this in Matlab. Matlab is 10-100 times slower than C or Fortran.
 
Thanks for the replies. I'll look into Gaussian quadrature.

I feel silly not having thought of the symmetry of the curve, since I know exactly where the centre is already. Unfortunately I don't have access to C or Fortran, and my code needs to integrate with a lot of other code I have already written in MATLAB.

I didn't realize it was so slow though? How can it be so slow?
 
You most certainly do have access to C. C/C++ comes bundled with the system if you are working on a Linux machine or a Mac. If you are working on a Windows machine, there are plenty of free C/C++ compilers available, some directly from Microsoft.

Matlab is a nice tool for developing an initial solution to some problem. It's graphics are phenomenal. It has huge libraries of tools and techniques. However, it is slow. It is an interpreted language.

Since you are using Matlab, learn to use the huge set of tools and techniques that are an integral part of Matlab. Look into Matlab's numerical quadrature routines. Matlab supplies a Gaussian quadrature function, and it is rather smart (adaptive). You do not need to reinvent the wheel here; besides, your wheel won't look as nice as Matlab's wheel. All you need to supply is the function and the integration interval to Matlab's Gaussian quadrature function -- and this can be an infinite interval with that function.
 
Insights auto threads is broken atm, so I'm manually creating these for new Insight articles. In Dirac’s Principles of Quantum Mechanics published in 1930 he introduced a “convenient notation” he referred to as a “delta function” which he treated as a continuum analog to the discrete Kronecker delta. The Kronecker delta is simply the indexed components of the identity operator in matrix algebra Source: https://www.physicsforums.com/insights/what-exactly-is-diracs-delta-function/ by...
Suppose ,instead of the usual x,y coordinate system with an I basis vector along the x -axis and a corresponding j basis vector along the y-axis we instead have a different pair of basis vectors ,call them e and f along their respective axes. I have seen that this is an important subject in maths My question is what physical applications does such a model apply to? I am asking here because I have devoted quite a lot of time in the past to understanding convectors and the dual...
Thread 'Imaginary Pythagoras'
I posted this in the Lame Math thread, but it's got me thinking. Is there any validity to this? Or is it really just a mathematical trick? Naively, I see that i2 + plus 12 does equal zero2. But does this have a meaning? I know one can treat the imaginary number line as just another axis like the reals, but does that mean this does represent a triangle in the complex plane with a hypotenuse of length zero? Ibix offered a rendering of the diagram using what I assume is matrix* notation...
Back
Top