Hi,

considering a report made by omorr, I found this bug:



here is the file in the image: howisthisposible_debug.zip


regards,

w3b5urf3r

Thanks to both of you w3b5urf3r and Oscar for investigating this issue. It seems something is really problematic in SMath regarding all of this.

I am surprised with this one - definitely a bug

BTW., as we mentioned in the previous post, if you use [MATH=eng]f(X):eval(transpose(X)*X)[/MATH] or [MATH=eng]f(X):line(res:eval(transpose(X)*X),res,2,1)[/MATH] you can get the right answer, regardless of Optimization of course. The problem is that you can not use eval() sometimes. It will simply not work. It would be very useful if we could use eval() every time when we need just the numerical answer. But I think that this is not possible due to the internal SMath engine. To be honest, as I mentioned few times, I really do not know when we can use eval() and when we can not. It is a bit trial-and-error procedure for me.

Hope that Andrey will have the solution to these problems ASAP.

Regards,
Radovan

Wrote

Hi,

considering a report made by omorr, I found this bug:



here is the file in the image: howisthisposible_debug.zip


regards,

w3b5urf3r

I investigated a little bit more and I found that there are relations between optimization type applied to function definition and function evaluation. With numeric optimization and none optimization, the results are right regardless optimization type applied to function evaluation:



Oscar Campo

This is because program performs simplification of the X^T*X - for some reason SMath Studio thinks that X*X^T is more readable or that it will be calculated faster (just guessing, any reason could be here). And program have no idea what is X - matrix or not. So it looks like correct behavior. I don't know, but I think other programs should have the same issue. Obviously it should happen only with symbolic optimization.

Fixed.

Best regards.

---

SyLib.7z (25 KiB) downloaded 49 time(s).

It seems it is working OK now, thank you Andrey

Regards,
Radovan