3 ms·
I've always thought it would be helpful to have floating point values that tracked the upper and lower bounds of the possible error. That would make analysis of
by mark-r 13y ago
I've always thought it would be helpful to have floating point values that tracked the upper and lower bounds of the possible error. That would make analysis of the results much easier.
- robzyb 13y agoBut what about the accuracy of the error bounds?
- pascal_cuoq 13y agoIt is easy to track the real result of the computation with floating-point intervals: round down when computing the lower bound, and up when computing the upper bound of each interval.
- jordigh 13y agoThat exists and is called interval arithmetic: https://en.wikipedia.org/wiki/Interval_analysis https://en.wikipedia.org/wiki/Interval_analysis
- wbl 13y agoThat's called ball arithmetic, and there are many libraries for it, such as arb. http://fredrikj.net/arb/ http://fredrikj.net/arb/ It's useful in mathematical calculation, but usually with floating point calculations you have some error bounds coming from your algorithm which are "good enough" if the result is going to be the value of something with 5% tolerance.