TVirtualFitter: Problem fixing a minimum step size for my parameters

[LeoB]

Hi,
I’m am trying to fit some datapoints with 5 different curves using TVirtualFitter and fcn().
I basically have 6 different parameters to vary. The first 5 are only scalings parameter between 0 and 1 most of the time, no problem here. But the 6th one is different and changes the shape of one of my curves. That mean everytime I want to change it, I have to recalculate every points of the curve and it takes a lot of time. Fortunately I don’t need a lot of precision on that parameter. it can only vary between 10 and a 100 and step of 0.5 would already be great.
The problem is that Root make it vary to a precision of 1e-6 or more, which I don’t need and takes to much time.
So is there a way to fix this step for only one parameter?
I tried using the TVirtualFitter::SetParameter(ipar, parname, value, verr, vlow, vhigh), but it doesn’t seem to have any effect.
It would seem to be a basic functionality, but I couldn’t find it.

Thanks for your answers,
Leo

[moneta]

Hi,

During the minimisation you cannot control the step sizes. These are set internally in order to compute the derivative for the parameter. A possible solution would be to provide an external gradient to the minimiser, but then you would need to provide for all the parameters.

Best Regards

Lorenzo

[LeoB]

OK thanks, I’m already trying to do the iteration by hand…