Re: GSoC 2017 - Improve iterative methods for sparse linear systems

[Prabhanjan Mannari]

On Fri, Mar 3, 2017 at 9:17 PM Marco Caliari <address@hidden> wrote:

On Fri, 3 Mar 2017, Prabhanjan Mannari wrote:

>  Hello
> I am Prabhanjan Mannari, and I'm interested in improving iterative methods for sparse linear systems [1]
>
> I have gone through the project description, and I have a few questions-
>
> 1. What are the pre-requisites that I should have to take up the project ?
> 2. Is the knowledge of the entire reference book [2] listed in the project description necessary for the implementation of the project (or are a few chapters more relevant than the others) ?
>
> [1]: http://wiki.octave.org/Summer_of_Code_Project_Ideas#Improve_iterative_methods_for_sparse_linear_systems
> [2]: www-users.cs.umn.edu/~saad/IterMethBook_2ndEd.pdf

Dear Prabhanjan,

you need to know some linear algebra and m-file programming, at least. For
instance, can you implement the standard conjugate gradient method in GNU
Octave (algorithm 6.18 of the book?). Then, cloning GNU Octave primary
archive and compile it from the sources. Have you already tried? Finally,
can you apply the patch developed by Cristiano last year and understand
his work?

About the book, at least four additional functions are missing in GNU
Octave. Therefore, you need the chapters relevant for these functions.

Marco

Dear Marco

I have attached an implementation of the conjugate gradient method. I went through only the specifics of the conjugate gradient method to implement the algorithm. I'm yet to study the method in the Krylov subspace framework, and its convergence analysis. I'm open to comments and suggestions on how to improve the code.

I'm comfortable with building octave from the sources on a Linux system. I have not gone through the patch developed by Cristiano yet. I feel a more thorough understanding of the methods(functions) is required before I can understand his work.

Prabhanjan

 

conjugate_gradient.m
Description: Text Data