Fedora Linux Support Community & Resources Center
  #1  
Old 7th March 2008, 04:27 AM
jkcpavan Offline
Registered User
 
Join Date: Feb 2008
Location: India
Age: 29
Posts: 9
Exclamation parallel matrix inversion



Hai every body

i want to parallalise the computation of matrix inversion. can you suggest which is the best technique when we think in parallel. if you are having any pseudo code please explain me.
__________________
Learn-->Discover-->Develop------------>Share
Reply With Quote
  #2  
Old 7th March 2008, 08:43 AM
bmvbab Offline
Registered User
 
Join Date: Jun 2005
Location: India
Age: 33
Posts: 376
just give a look in freshmeat.net under
scientific/engineering->mathematics
Reply With Quote
  #3  
Old 7th March 2008, 08:51 AM
jkcpavan Offline
Registered User
 
Join Date: Feb 2008
Location: India
Age: 29
Posts: 9
thnk you for your suggestion boss

but it is not having any information about inversion ..... the site contains the basic matrix operations but i want to to the matrix inversion that too parallely
__________________
Learn-->Discover-->Develop------------>Share
Reply With Quote
  #4  
Old 7th March 2008, 11:17 AM
bmvbab Offline
Registered User
 
Join Date: Jun 2005
Location: India
Age: 33
Posts: 376
well then how about this:
http://portal.acm.org/citation.cfm?id=321434
Reply With Quote
  #5  
Old 7th March 2008, 11:48 AM
jkcpavan Offline
Registered User
 
Join Date: Feb 2008
Location: India
Age: 29
Posts: 9
sorry to say this boss....

actually i am having account in acm portal but iam not an authorised to user to view the document you hav sent. thank you for helping me.

if possible can you send me the document you have shown.
__________________
Learn-->Discover-->Develop------------>Share
Reply With Quote
  #6  
Old 18th March 2008, 11:41 PM
Lascolb Offline
Registered User
 
Join Date: Dec 2007
Posts: 233
You cant view the PDF file,
Or is thr another file your talking about?

In any case you shouldn't need to be a authorised user to view the document.
Reply With Quote
  #7  
Old 19th March 2008, 10:36 AM
aleph Offline
Banned (for/from) behaving just like everybody else!
 
Join Date: Jul 2007
Location: Nanjing, China
Posts: 1,332
I think, in general you don't need code to do matrix inversion very often. We need matrix inversion to solve linear systems, and writing an algorithm for solving, for example, y = Ax, is way much more efficient than first finding A^(-1) then calculate A^(-1)y, because we waste many computrons doing the boring multiplication in the latter step which can often be merged into the solving step. Most of the time we do this by decomposing the matrix into certain factors, e.g. lower and upper triangular, SVD, etc ... and solve the equation stage by stage, (approximately) eliminating one factor after one. At least, that's common for sequential computation.

I don't know what things will look like in the fields of parallel computation. I guess there is no unified answer either coz there are too many types of matrices...

There are quite a lot of books on the topic... If you do a search on the web you get pretty many hints. As for the ACM paper, are you having an account for your library or yourself? I have a library access and there's no problem downloading... I don't even have to login.
__________________
Code:
from rlyeh import cthulhu
cthulhu.fhtagn()

Last edited by aleph; 19th March 2008 at 10:41 AM. Reason: typo
Reply With Quote
  #8  
Old 19th March 2008, 05:33 PM
Magnar Offline
Registered User
 
Join Date: Dec 2007
Posts: 68
One of my teachers, who is good at computational science, said that Jacobi's method for diagonalisation has become more popular lately because it is pretty easy to parallellize.
Reply With Quote
  #9  
Old 25th March 2008, 05:24 AM
jkcpavan Offline
Registered User
 
Join Date: Feb 2008
Location: India
Age: 29
Posts: 9
hello every body

Mr. aleph is saying that solving the linear equation is better method. but i think it may not be that much better. i think jacobi mettod is better. but the main problem i am getting is that wheather the equation i am entering to fid the solution/inverse would lead to result. i.e the solution exists or not. for that please suggest me....
__________________
Learn-->Discover-->Develop------------>Share
Reply With Quote
  #10  
Old 27th March 2008, 09:48 PM
Magnar Offline
Registered User
 
Join Date: Dec 2007
Posts: 68
Can't you just calculate the determinant, to see if the matrix is invertible (if solutions exist)?
Reply With Quote
  #11  
Old 28th March 2008, 04:32 PM
aleph Offline
Banned (for/from) behaving just like everybody else!
 
Join Date: Jul 2007
Location: Nanjing, China
Posts: 1,332
Quote:
Originally Posted by Magnar
Can't you just calculate the determinant, to see if the matrix is invertible (if solutions exist)?
This is way too expensive ...

Generally, you can only estimate whether the solution exists. Finding an approximation of the condition number may be a good start.
__________________
Code:
from rlyeh import cthulhu
cthulhu.fhtagn()
Reply With Quote
  #12  
Old 2nd April 2008, 02:33 PM
jkcpavan Offline
Registered User
 
Join Date: Feb 2008
Location: India
Age: 29
Posts: 9
hi to alla caliculating determinent is the most difficult part if that is done we can directly use that technique but our aim is use some simple technique... what is ment by approximation and how to caliculate that one please explain me
__________________
Learn-->Discover-->Develop------------>Share
Reply With Quote
  #13  
Old 2nd April 2008, 03:32 PM
MarkS Offline
Registered User
 
Join Date: Feb 2008
Posts: 53
jkcpavan, it may be difficult get an answer your question here as it is not related to fedora or even linux. Why not try an mpi, or openmp, or hpf forum? Lots of math, numerical methods people hang out there.
You also need to be a bit more specific;
- How big is that matrix anyway? Todays cpus are fast, communication between them, not so fast.
- What assumptions can you make about the matrix?
- If you don't know if a solution exists that may be weird (or not) depending on your area of study.
- Can you decompose the matrix in any useful way?
- Why is calculating the determinant difficult? Slow maybe, but difficult?
Reply With Quote
  #14  
Old 3rd April 2008, 11:50 AM
jkcpavan Offline
Registered User
 
Join Date: Feb 2008
Location: India
Age: 29
Posts: 9
no it is not difficult to answer this question it is all about the programming not about the parlalization ofcourse parlalization is a good one it is a different thing but what i am requesting is the corectness, scalability and usability of code.

here i am having a link where the matrix inversion is explained clearly.. but i am not getting right can any body read and explain the algorithem

http://oddprogrammers.blogspot.com/

please reply me regauring this code on the site so that ill try to understand
__________________
Learn-->Discover-->Develop------------>Share
Reply With Quote
  #15  
Old 5th April 2008, 10:23 AM
pal_thomas2009 Offline
Registered User
 
Join Date: Apr 2008
Posts: 1
Mr jkcpavan is saying that he had found of matrix inversion in gaussmethod then what will be your problem, if you solve the equations if the solutions does not exist also it will give you the result it is the power of the gauss jordon method i think. so, no need to worry about that and also i am trying to parlalizing the matrix inversion using the same method after visiting your suggested site www.oddprogrammers.blogspot.com it is good to see such explanation but the main problem in that algorith i found it is using two nore temporaryn arrays. it is good to look at memory complexity also. so explain me any body if they had done any work regaurding this
Reply With Quote
Reply

Tags
inversion, matrix, parallel

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
dot matrix printer changedtothis Hardware & Laptops 7 20th August 2009 04:54 AM
Matrix revisited Toyler Installation, Upgrades and Live Media 19 9th August 2005 03:08 PM
it's like the matrix... Toyler Installation, Upgrades and Live Media 8 6th August 2005 07:19 PM
linux is like the matrix! DeveloperMCT Using Fedora 8 15th July 2005 07:40 PM
printing problem with dot matrix dreamsshinu Using Fedora 3 15th July 2005 12:51 AM


Current GMT-time: 16:42 (Thursday, 30-10-2014)

TopSubscribe to XML RSS for all Threads in all ForumsFedoraForumDotOrg Archive
logo

All trademarks, and forum posts in this site are property of their respective owner(s).
FedoraForum.org is privately owned and is not directly sponsored by the Fedora Project or Red Hat, Inc.

Privacy Policy | Term of Use | Posting Guidelines | Archive | Contact Us | Founding Members

Powered by vBulletin Copyright 2000 - 2012, vBulletin Solutions, Inc.

FedoraForum is Powered by RedHat
Sonneberg Photos - Celso Ramos - Takanosu