On Fri, Sep 14, 2007 at 08:07:51PM +0000, amli at comcast.net wrote:
> Hello,
>
> Thank you all for responding. I still have a question regarding
> numerical evaluation. This is part of a longer symbolic expressions
> which I am satisfied with. I am not just trying to evaluate dot
> product (which in the case of symbolic expression can be verified to
> show that dot product is zero-as shown in the input). When I remove
> keepfloat:true I get very large numbers with exponents 10^. I am
> computing my symbolic expressions for some particular cases. Am I
> losing precision when I see these large numbers due to ratsubst .
> My numbers seems to be very manageable from numerical point of view.
> Can I trust Maxima for numerical evaluations to back up my symbolic
> expresions? Thank you.
Why not evaluate the symbolic expression first, and then plug in your
numerical values at the end?
However, as others have said, once you have converted to rationals,
there is no longer any round off error in further computations.
You can control the precision of the conversion to rationals by
setting ratepsilon, which the docs say is 2e-8 by default. (so
abs(rat(f)-f)/f < 2e-8).
--
Daniel Lakeland
dlakelan at street-artists.org
http://www.street-artists.org/~dlakelan