RE: Condition/ constaint problem
- To: mathgroup at smc.vnet.net
- Subject: [mg41060] RE: Condition/ constaint problem
- From: "Ersek, Ted R" <ErsekTR at navair.navy.mil>
- Date: Wed, 30 Apr 2003 04:24:29 -0400 (EDT)
- Sender: owner-wri-mathgroup at wolfram.com
> In David Withoff's earlier message that I refer to below, he > suggested that maybe future > versions of Mathematica shouldn't do numerical derivatives because > it can't always do it reliably. > I think it would be good to allow for that as an option. Something > like > SetOptions[Derivative, Method->None] > would ensure Derivatives are not approximated. While > SetOptions[Derivative, Method->EulerSum] > would work like the default implementation of > NumericalMath`NLimit'ND. > > ----------------------------- > > At > http://forums.wolfram.com/mathgroup/archive/2003/Apr/msg00688.html > David Withoff explained how NumericalMath`NLimit`ND can be used to > correct for a problem with approximating a derivative via the built-in > Derivative function. David said the built in Derivative function does the > same thing as NumericalMath`NLimit`ND but without knowing the scale of > variation of the function. Perhaps the built in Derivative function > should have the same options as NumericalMath`NLimit`ND which are > (WorkingPrecision, Scale, Terms, Method). Of course the setting of these > options would only be relavent when numerical differentiation is needed. > Also these options could have default settings that would cause the > calculation to work as it does now (ie. Scale->Automatic). > > Regards, > Ted Ersek > >