Версія даної теми для друку

Натисніть сюди для перегляду даної теми у оригінальному форматі

Розподілені обчислення в Україні _ Мікс - Інші завершені проекти _ Evo@home

Автор: nikelong Jan 24 2011, 11:42



Проект "Evo@home"

----------------------------------------------------------------------------------------------------------

----------------------------------------------------------------------------------------------------------
ТОП-20 участников:
[img_]картинка[/img_]
----------------------------------------------------------------------------------------------------------
Дата основания команды - 22.11.2010 Капитан - distributed.org.ua
----------------------------------------------------------------------------------------------------------
Для присоединения к команде Украины:
1. Загрузите http://boinc.berkeley.edu/download.php (Если его у Вас еще нет!)
2. Перейдите в "расширенный вид"
3. Выберите сервис ---> добавить проект
4. Введите адрес проекта http://boinc.run.montefiore.ulg.ac.be/evo/
5. Введите свои регистрационные данные.
6. Найдите нашу команду. Она называется Ukraine и адрес ее http://boinc.run.montefiore.ulg.ac.be/evo/team_members.php?teamid=476&sort_by=total_credit&offset=0 вы могли видеть выше.
7. Если есть доступные для загрузки задания Вы их получите и начнете расчеты.
----------------------------------------------------------------------------------------------------------
Новичкам: http://distributed.org.ua/index.php?go=Pages&in=view&id=170, как поставить и настроить BOINC-менеджер
----------------------------------------------------------------------------------------------------------
Полезная информация:
Для идентификации пользователя в BOINC могут служить 2 вещи:
1) пара e-mail/пароль
2) межпроектный идентификационный ID (Cross-project ID) - 32значное шестнадцатиричное число.

Если Вы пожелаете подключится ещё и к другому BOINC-проекту, то помните: чтобы не плодить новых аккаунтов при подключении к новому проекту или команде, нужно обязательно везде регистрироваться с одним и тем же Именем и EMAIL. если при регистрации в проекте указать другой e-mail , BOINC создаст новый аккаунт с тем же именем! В этом случае рекомендуется зайти во все ваши аккаунты и во все проекты и где надо поменять емейл на нужный. Через некоторое время ваши аккаунты сольются в один с одним cross-project-id.
----------------------------------------------------------------------------------------------------------


О проекте:
Evo@home is a research project that uses Internet-connected computers to do research in Machine Learning. This project uses Evolutionary Algorithms to optimize the parameters of different kind of machine learning algorithms.
Currently the project is used to tune the parameters of a learner used for the prediction of the three-dimensional structure of a protein from its sequence of amino acid residues. You can participate by downloading and running a free program on your computer.
Evo@home is based at Montefiore. This is the department of Electrical Engineering and Computer Science of the University of Liиge (Belgium).
More information about our research will be available soon.

Ссылки по теме:График ППД команды "Ukraine"

Автор: Death Jan 24 2011, 14:02

I'm in.

Автор: re_SET Jan 24 2011, 21:58

So, let's crunching! koc.gif

Автор: Rilian Jan 26 2011, 16:40

2011-01-23: I'm glad to announce that I just turned ON the workgenerator. It will be on during a few weeks while we are working on the final application. The application used currently is still the "baby" application. The goals of this step of the project are: - See if the modifications done to the workgenerator are useful (avoid too quick convergence of the evolutionary algorithm). - Test the stability of the server (the project is now public). The computing time of a WU is on average between 10 and 40 min. The display of the fraction done (progress in %) is NOT working yet, but the application should be working fine. The memory usage (RAM) should be less than 250Mo. I'd like to thanks Laurent Schoonbrodt for the design of our logo (http://www.laurent-s.com/). Thank you for your support.

Автор: Rilian Jan 27 2011, 01:16

Evo@home: Already successful results!
2011-01-26: Thanks to your help, I'm glad to announce that we have already made some really interesting discoveries! Let me tell you more about it. The WU's processed at the moment are used to build and find the best model for the prediction of the secondary structure of proteins. Before using BOINC, this tuning has been realized "by hand" by researchers. The best score was around 78%. Before opening this project, I ran some beta-tests on a private Boinc project with some members of l'Alliance Francophone. In two days we reached a score of 79%. During this week, thanks to your help, we reached a surprisingly high score of 86%. Unfortunately, this score is too high to be real. Indeed, it turns out that we found a bug in the evaluation procedure. We have also discovered some new pathologic cases. To sum up, by calculating for our project this week, you have helped to: - Prove the efficiency of the Evolutionary Algorithm used (optimizer), as the mean of the scores keeps increasing every day! - Find an important bug in the evaluation procedure (calculation of the score) leading to too high scores. - Find some pathologic cases leading to scores of 100% or 0%. During the next weeks we are gonna improve our application and correct the bugs. I'd like to state again that what your are calculating for at the moment is just the first step of our research program and this is why interesting results can be found so fast. We hope to release the full application as soon as possible. As we don't like wasting computation time, I turned off the work generator. There won't be any work for at least two weeks. After that there will be some batch of WU's for debug purpose, so please stay tuned. Thank you for your support!

Автор: Rilian Feb 8 2011, 14:40

2011-02-08: While you are waiting for new WU's I thought it could be useful to provide you some more information about this project. An article concerning our research will be presented in two months at a conference. Of course this article cannot be published before the conference but here is a poster presenting the subject: http://www.jul-1.be/download/P02.pdf Using BOINC and new optimizers, we hope to find even more accurate results! Concerning the source code of this project, it is now available here. This source code evolves constantly and it is distributed under GNU GPLv3. New WU's should be available by the end of this month.

Автор: Rilian Mar 1 2011, 12:25

Админ проекта пишет что с 7 марта будут "научные" ВЮ

CODE
http://boinc.run.montefiore.ulg.ac.be/evo/forum_thread.php?id=22&nowrap=true#124

QUOTE
By the way: Scientific WU's should be available by the 7 of March smile.gif

Автор: Death Mar 20 2011, 23:09

2011-03-20: As you already know, this project is memory intensive. We try to not use more than 1Go in practice and the theoretical limit is set to 1.5Go. If you haven't set your max RAM usage preferences and if you have a multi-cores computer, you computer might become out of control / freeze. For instance imagine you have a quad-core with 4Go of RAM, if 4 Evo's WUs are running simultaneously, you'll have no more RAM free...
During the next week, I'll set the maximum WUs in pogress on a host to 1 so that you have time to set your preferences. After that there won't be any limitation, so please don't forget to set you RAM usage preferences.

Автор: x3mEn Nov 9 2011, 10:51

Retired

Автор: nikelong Mar 17 2012, 23:03

(x3mEn @ Nov 9 2011, 10:51) *

Retired


А с чего взял? Только потому шо сайт лежит, или где то это было написано?

Автор: Rilian Mar 17 2012, 23:04

QUOTE(nikelong @ Mar 17 2012, 23:03) *

QUOTE(x3mEn @ Nov 9 2011, 10:51) *

Retired


А с чего взял? Только потому шо сайт лежит, или где то это было написано?

на сайте написано что он в дауне уже 2й год

Invision Power Board
© Invision Power Services