Ublas :

Subject: Re: [ublas] whole ublas matrix as vector
From: Karl Meerbergen (karl.meerbergen_at_[hidden])
Date: 2009-09-07 10:05:58

Michael Norel wrote:
> Hi All.
>
> i tried to use "norm_inf" for matrices but i get unexpected result
>
> //-------------------------------------------
> Matrix m(2, 3);
> Vector v(2*3);
>
> for (SizeType i = 0; i < m.size1 (); ++ i)
> {
> for (SizeType j = 0; j < m.size2 (); ++ j)
> {
> v (i * m.size2() + j) = m (i, j) = -(Real)(3 * i + j);
> }
> }
>
> std::cout << m << std::endl;
> std::cout << v << std::endl;
>
> std::cout << norm_inf(m) << std::endl;
> std::cout << norm_inf(v) << std::endl;
> //-------------------------------------------
>
> .[2,3]((-0,-1,-2),(-3,-4,-5))
> [6](-0,-1,-2,-3,-4,-5)
> 12
> 5
>
> Is it expected result? what is math meaning of norm_inf(m) ?
>
> Thanks
>

This is correct: 0+1+2=3 and 3+4+5=12. The maximum of both is 12.
See http://en.wikipedia.org/wiki/Matrix_norm

Karl

>
> B
>
> --- On Sun, 9/6/09, Gunter Winkler <guwi17_at_[hidden]> wrote:
>
>
>> From: Gunter Winkler <guwi17_at_[hidden]>
>> Subject: Re: [ublas] whole ublas matrix as vector
>> To: "ublas mailing list" <ublas_at_[hidden]>
>> Date: Sunday, September 6, 2009, 9:11 PM
>> Michael Norel schrieb:
>>
>>> Hi all!
>>>
>>> I still need this functionality (matrix expression as
>>>
>> vector expression)for intensive use of ublas.
>>
>>>
>>>
>> Please check out the latest development version from SVN:
>> see

>> and have a look at
>> experimental/sparse_view.hpp, class c_array_view. This
>> class should
>> fulfill the immutable part of storage array concept. Now
>> only a class
>> fulfilling the immutable part of the vector concept must be
>> developed
>> which can be used to present any storage array as a read
>> only vector.
>>
>>
>>> I need to
>>> 1. compare 2 matrixes with some eplison threshold like
>>>
>> "norm_2(m1-m2) < eps"
>>
>>>
>>>
>> Why not using norm_inf or norm_1 for matrices?
>>
>>
>>> 2. check if the submatrix of matrix is zero like
>>>
>> "norm_2(m1) > 0"
>>
>>>
>>>
>> I'd also suggest to use norm_inf or even better a customer
>> function that
>> aborts the check at the first non-zero element.
>>
>>> 3. find the absolute maximum element of matrix like
>>>
>> norm_inf(m1)
>>
>>> and many others, where m1 and m2 is matrix
>>>
>> expressions.
>>
>>>
>>>
>> Could you prepare a list of the most used matrix functions?
>> Maybe there
>> are others interested. Then we could easily add a
>> which contains such methods.
>>
>> There is a trac for boost: https://svn.boost.org/trac/boost/
>>
>> Here you can view, comment and create new tickets. Feel
>> free to create a
>> feature request for uBLAS. Try to add an example or
>> pseudo-code to
>> express the expected usage.
>>
>>
>>> Where can a i post the functionality request for such
>>>
>> adaptor? Or it is realy the right way to implement it
>> myself?
>>
>>>
>>>
>> The quickest way is always do it on yourself and to post
>> the patch, an
>> example and a unittest here. Then I can review the code and
>> commit it to
>> the repository.
>> _______________________________________________
>> ublas mailing list
>> ublas_at_[hidden]

>> Sent to: minorlogic_at_[hidden]
>>
>>
>
>
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]

> Sent to: karl.meerbergen_at_[hidden]
>

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm