symb, num, none - quite a difference!

symb, num, none - quite a difference! - 消息

#1 已发布: 2012/2/11 16:57:35
Radovan Omorjan

Radovan Omorjan

325个赞和1个踩,共2,052个帖子。

组: Moderator

Hello,

Here is a part of the file and the expression on the left is Optimization|Symbolic



Quite different numbers than the Numeric and None





Here is the entire file Primer29.sm

This time the same is on both WinXP and Ubuntu. If nobody can reproduce this one as well () - something must be quite wrong with my computer.

Regards,
Radovan
When Sisyphus climbed to the top of a hill, they said: "Wrong boulder!"
#2 已发布: 2012/2/12 07:22:13
Andrey Ivashov

Andrey Ivashov

2,344个赞,共4,057个帖子。

组: Super Administrator

Reproduced. Will check it later today. Thank you.

Best regards.
1位用户喜欢此帖子
Radovan Omorjan 2012/2/12 07:44:00
#3 已发布: 2012/2/12 13:17:52
Radovan Omorjan

Radovan Omorjan

325个赞和1个踩,共2,052个帖子。

组: Moderator

Thank you Andrey,

Here is another part from one of my working files. Difference again between Symbolic and Numeric optimization of the equations.

Symbolic:

The accompanied file sym.sm

Numeric:

The accompanied file num.sm

Please look at the yellow regions. I am not quite sure which one is correct, but solving in another software gave the same solution when the SMath numerical optimization was used. Moreover, further calculation showed that the numerical result should be accepted and the symbolic is questionable.

To be honest, I am a bit frustrated that roots() failed here as well. If it works for you - than must be my computer again.
There were another situations when roots() failed in another of my working files. I will post them as soon as I get them ready to report the problem.

Here is a file where roots() failed for me in every calculation (either I made a mistake I didn't see or something else) Primer29-roots%20failed.sm. Fortunately, Newton-Raphson worked here and I managed to solve this somehow. Playing with Optimization will make a mess - Jacobian elements equal to all zeros, different results etc.

I desperately need a rock solid numerical solvers for nonlinear system of equations built in SMath but, unfortunately, roots() disappointed me at the moment.

Regards,
Radovan
When Sisyphus climbed to the top of a hill, they said: "Wrong boulder!"
#4 已发布: 2012/2/13 05:08:40
Infinity

Infinity

3个赞,共30个帖子。

组: User

Hello,

I found also a problem when calculating with numeric optimization.
The following simple calculation adds up some integer numbers with floating point numbers.
The result should always be zero. The first numeric solution does not equal zero. When I introduce some brackets around the floating point numbers, the result is correct.
When using symbolic optimization the result is always correct.

Left: Numeric, Right: Symbolic


File:
optimization.sm

  • 有新帖
  • 无新帖