Error in simple calculations

2015 4 17
1 个回答
更新时间：2015 4 18
3 次查看（30 天）

0 个投票

Hi everyone I'm really having a very weird and silly problem with Matlab. A very simple calculation which is z=(x-y). That's all, but when x= 0.327999967200000 and y=0.319999968000000 then z=0.00799999920000000 which is correct. But when x=0.311999968800000 and y=0.303999969600000 then z=0.00799999920000005 which is wrong as the answer should stay 0.00799999920000000. It is a simple silly stupid operation nevertheless the Matlab is unable to solve it correctly. I used everything, the format the cast the eps() non of them worked and the error is still persistent. What shall I do so the Matlab would perform such calculation without errors. Is it a bug or something? I'm using Matlab 2014b.
Thanx in advanced

请先登录，再回答此问题。

回答（1 个）

Sebastian Castro
Sebastian Castro 2015-4-17

0 个投票

Definitely not a bug -- just floating-point error: http://www.mathworks.com/matlabcentral/answers/57444-faq-why-is-0-3-0-2-0-1-not-equal-to-zero
If you care about that level of precision, it might be feasible to work with fixed-point data. This way, you can choose a good fixed-point implementation that can exactly represent a particular set of numbers.
- Sebastian

1 个评论
显示 -1更早的评论 隐藏 -1更早的评论

Ahmed
Ahmed 2015-4-18
Dear Sebastian I'm really grateful for the answer but I wonder if there is a simpler way to do it. Actually these number are indecies for voxel cells and they are in millions multiply that by 8 for each cube. Using such approach yields the execution time to explode. I want to avoid this. Besides, the geniuses and huge minds that created such sophisticated scientific package should address and fix this annoying problem. Any calculator, windows calc or a free calc, can perform such operation flawlessly. Also, we as user should have full confidence in Matlab's results not wondering if there was a bug or round-off error may cause the results to be like this. I know it is a precision and word length problem but honostly I saw this for the first time with Matlab and I'm a C++ user.
Any way is there another approach less time consuming and more direct to solve it? I would appreciate it.
Thanx alot

请先登录，再进行评论。

类别

在 帮助中心 和 File Exchange 中查找有关 Logical 的更多信息

标签

提问:

Ahmed
Ahmed
2015-4-17

评论：

Ahmed
Ahmed
2015-4-18

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by