02-26-2009 10:53 PM
For example
ten points arranged in a parabola,
is there a function to predict the coordinates of the six point behind according to the four points in front of?
02-27-2009 02:17 AM
You could use a curve-fitting algorithm to determine a function and then calculate the new coordinates...
hope this helps,
Norbert
02-27-2009 02:32 AM
Hi,
As Norbert said you can find some good examples of the Levenberg-Marquardt algorithm in the example finder of LV
Regards,
02-27-2009 11:21 PM
The ability to predict data values is predicated on your data having some sort of inherent structure. In cases where your data has that structure, curve fits can be very helpful. However, if your data does not have any inherent structure, or is too noisy to be able to see the structure, you can run curve-fits and still get results that look reasonable. Unfortunately those results would probably correlate better with the phase of the moon than they would what you are measuring. No matter how fancy, mathematics can only predict a probable value based on what it was given - and as in all things computer: GIGO. Hence, when working with this sort of thing, I always bear in mind this definition of a pocket calculator:
"...a device that can divide a ball park estimate by a wild-ass guess and come up with an answer accurate to 13 decimal places..."
Mike...
03-01-2009 10:42 PM