[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: GSOC16 - Improve iterative methods for sparse linear systems
From: |
Cris |
Subject: |
Re: GSOC16 - Improve iterative methods for sparse linear systems |
Date: |
Tue, 15 Mar 2016 10:28:06 +0100 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.6.0 |
On 11/03/2016 14:34, Marco Caliari wrote:
On Tue, 8 Mar 2016, Cris wrote:
Dear all,
I'm Cristiano Dorigo, a 2-nd year student of the Master Degree in
Mathematics of the University of Verona, Italy.
I'm interested in partecipating at the GSOC 2016 to work at the
project "Improve iterative methods for sparse linear systems".
I have a good knowledge of Octave and Matlab built during the
Bachelor in Applied Mathematics through some exams (Numerical
Analysis, Numerical Methods for Differential Equations, Fluid
Dynamics) and also during the Master Degree. In particular I followed
a course on Spline theory and another on the Finite Elements Method
with a part fully reserved to iterative methods for sparse linear
systems.
I also know the Java code language from the Computer Programming
course at the Bachelor.
I'm interested in this project because during my Bachelor Thesis I
studied the Krylov subspaces, the Arnoldi Method and the GMRES
algorithm. In particular I implemented (using Octave) the GMRES
algorithm using the Householder orthogonalization instead of the
Gram-Schmidt one and studied the quality of the solutions with the
GMRES with these different orthogonalizations. I think that with this
project I can study in deep the topics that I already faced in the
Thesis and I can study other algorithms close to the GMRES, in such a
way to improve my general knowlegde about sparse linear systems
algorithms.
Don't hesitate to ask anything if there are questions about my
student career or about my thesis.
Dear Cristiano,
so you already implemented from scratch a function like gmres.
Therefore, you may have a rough idea of the required effort for the
project. Can you make a more detailed scheduling of the work, in case
the project is accepted? Please take honestly into account also your
examination session, if any.
Cheers,
Marco
Dear Marco,
thanks for your reply.
I think that the first thing that must be done for this project is to
study the codes that need an improvement (pcg, pcr, bicg, bicgstab, cgs,
gmres and qmr) and where this is necessary. I think that this first part
needs more or less 1-2 weeks for the study part and the same for the
improve part.
After this first part, I will try to implement the other algorithms
mentioned in the project one by one, and I can't give a precise
timetable for this part, but I can tell how I will work: I study the
algorithm and then I try to implement until it's done. Then I'll pass to
the next and so on. In this way I hope to implement as much algorithms
as I can.
My examination session begins on 15-th June and ends on 30-th of July,
and now I don't know precisely the dates of the exams. But to do not
take so much time from the GSOC, I will study for no more than 2 exams
and to stop the work for about 20 days in total (and I hope to organize
these days in such a way they are not 20 consecutive, but in 2 blocks of
10 days). If the project was accepted, when I'll know the exam's dates
(around the half of May), I write to the mentors to organize precisely
the timetable during the session.
Best regards,
Cristiano Dorigo