On Sep 5, 4:43 pm, Jason Grout <jason-s...@creativetrax.com> wrote: > On 9/5/11 5:40 PM, rjf wrote: > > > One relatively clean resolution of some of the problems > > is to replace every float -- on input -- to > > an equivalent rational. > > It seems like we've had some big problems with this, in that maxima will > sometimes simplify things because it assumes an exact rational number > was entered, when in reality, an approximate number was entered [1].
Is a floating point "input" number an approximate number, or an exact rational number that is expressed in a decimal form? Most reasonable numerical analysis (exception: Mathematica's significance arithmetic, which is not, in my opinion, reasonable), starts from the view that a floating point representation is fraction X base^exponent. Not (fraction +-slush) X base^exponent. > Maybe a more correct "clean" resolution is to convert every float to a > rational interval, or a float interval, and do interval arithmetic? no, a float is NOT an interval. it is a number. Of > course, that can get really nasty too... Presumably you must support intervals, since some of the M's do so, and Sage is supposed to be an alternative to them... RJF -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
