Gerasim@Home新項目 參考囉(俄國)

包含 BURP、Cels@Home、Chess960@Home、ESEA@home、Gerasim@Home、NCSSM Grid Computing、Pirates@Home、Project Neuron、RND@home 這些專案
回覆文章
alexpon
單核心電腦
單核心電腦
文章: 597
註冊時間: 2006-05-10 21:19
newboincstats_no: 21591
CPU 主力專案: 其他專案
來自: Taiwan

Gerasim@Home新項目 參考囉(俄國)

文章 alexpon » 2007-07-08 08:47

http://www.gerasim.boinc.ru/Gerasim/
Gerasim@Home
Thank you for your interest in Gerasim@home !Gerasim@home использует свободные вычислительные ресурсы Вашего компьютера. Своим участием Вы активно поддержите мировое научное сообщество.

隔了許久 2013年update 資訊 @@ 偷懶太久
引用:http://forum.gerasim.boinc.ru/default.aspx?g=posts&t=13 論壇資訊
Project science goal :
Testing algorithms for getting separations of parallel logic control algorithms.
google 翻譯: 測試算法的的並行邏輯控制算法得到的分離。
自己猜的意思: 平行邏輯控制演算法的 演算其separations ?
求 高手幫看要怎翻這句囉 :idea:
Detail: http://gerasim.boinc.ru/science/science.aspx
上方聯結 全俄文
下列google 翻的部份內容 有興趣 可以點進去看一看
研究領域
等價變換的圖形分區的的並行邏輯控制算法在應用程序中的邏輯控制系統的設計方案。
目前的科學目標
分析質量分區圖計劃的並行算法用於控制邏輯,通過各種試探。
項目介紹
邏輯控制系統(SLE)有最近成為更廣泛的因的增加使用的機器人自動化組裝線能夠與很少或根本沒有人類的干預(如參與或降低到最低限度),一方面進行必要的操作,和微電子技術的成功,穩步摩爾定律之後,對其他。這些系統的主要應用領域是:

•管理的重要組成部分錯誤時所造成的人為錯誤或任何其他原因,可能會導致嚴重後果的技術,經濟或社會性質(工廠管理,導彈,衛星等)。對於這一類的系統,特殊要求的高性能,容錯,如果可能的話,硬件複雜度低(因為其影響的大小和重量參數)和成本。
•管理的電子設備(處理器,加速器,專門計算器,等),其頻率可高達幾GHz。在這個角度來看,這種控制系統,高速和低成本擺在首位的特殊要求,部分原因是由於硬件的複雜度適中。
•控制與數字控制的機械手輸送線或機器,它本質上減少管理的慣性作動器不需要很高的性能脫穎而出的屈服準則的成本低,快速重新配置SLN的組成(新設備的操作,改變的可能性產品,等等)。

Application: separator ver.1.00


引用: http://forum.gerasim.boinc.ru/default.aspx?g=posts&t=15 論壇資訊
各階段 研究資訊 及BOINC 運算後之results 有興趣的可以點進去參考
Separator progress

大致簡介:
於 July 29, 2010 第一批實驗 有5700 WU 初步完成運算 也有成果資訊 請參閱內容
First 5700 WUs (experiment 1) was added. In this experiment we investigating dependencies between quality criteria, average algorithm size N (1-500) and restriction Xmax (3-40). Prior to completion of the calculation of all WUs new job will not be granted. At this moment has been complited calculation of 2250 WUs, will wait for others. First experiment results are presented here.

於July 30, 2010
Results of experiments was extended by a number of graphics showing the dependencies of probabilities \rho(\cdot) for separations with minimal value of quality criteria. Database with experimental results at this moment is 1.9 GB. Fast calculation of the new diagrams, which requires comparison of results from different files, as it turned out, is a nontrivial work.
At this moment has been complited calculation of 3334 WUs (58%), still will wait for others.

August 14, 2010
Calculation tasks for the first experiment was ended. Confirmed the advantage of parallel-sequential method for strong restrictions zone and S.I. Baranov method for weak restrictions zone. A number of new dependencies which appear with increasing N was discovered. A number of limiting factors was found (need additional experiments). Thanks to all volunteers who took part in the calculations!

August 14, 2010 第二階段 15100 WU
15100 WUs (experiment 2) was added. In this experiment we investigating dependencies between quality criteria, average algorithm size N (1-500) and restriction Wmax (3-60).

September 19, 2010
Postprocessing of results of second experiment was finished. Analysis of results revealed a number of features, which manifest themselves with increasing N. Parallel-sequential method as nice hand in strong (but not very strong!) and weak restriction zones by integral criterion J (behavior of partial quality criteria is individual, especially for H, X and \alpha); S.I. Baranov method is showing their best in weak restriction zone for small N (except for the criteria X — an advantage in a very strong restrictions zone and \alpha — an advantage both weak and strong restriction zones). Need additional study for weak restriction zone 60 < Wmax < 100, for what soon will be added WUs named wu_e2_*.wu. Thanks for all who was participated in the experiment!

September 27, 2010 第三階段 25872 WU
25872 WU for experiment 3 was added. This experiment studies the mutual influence of constraints 4 <= Xmax <= 25 and 3 <= Wmax <= 100 with different values of N \in {50, 100, 200, 300, 400, 500}.

October 17, 2010 第四階段 97000 WU
97000 WUs for experiment 4 was added. It is generally similar to the first experiment, but differs increased by 10 times the number of microoperations and logical conditions in the control algorithms and a range of changes of restrictions 4 <= Xmax <= 100 (in the future the range will probably be extended). Sample size increased to 5000 control algorithms. This will increase the execution time and outgoing traffic volume ~ 2.5 times by WU.

January 12, 2011 第五階段 150000 WU
The project is scheduled challenge named "The doggie of the Spanish breed" wich will be held from 20 till 27 january. More then 150000 WUs was added for new experiment 5: 3 <= Wmax <= 150, 1 <= N <= 600. In general, experiment 5 is similar to the previously finished experiment 2 but differs increased from 20 to 200 the number of microoperations and logic control signals. Good luck!

April 01, 2011 第五階段 運算產出
Results of experiment 5 was updated. Additional research in the field 151 <= Xmax <= 200 is required. More then 100000 WUs was added. During increasing the number of microoperations and logical control signals from 20 to 200 quality of separations obtained by S.I. Baranov method gets a little worse, parallel-sequential method is almost insensitive to change.
同日下午 第六階段 40萬WU 開始運算
400000 WUs for the sixth experiment was added: 3 <= Xmax <= 150, 4 <= Wmax <= 200, N in [50, 100, 200, 300, 400, 500, 600]. Sixth experiment is similar to the third experiment but differs from it increased from 20 to 200 the number of microoperations and logical controls and has a more wide range of the restrictions Xmax and Wmax.

August 01, 2011 某國出版書輯提及 但未多著墨 待專案運算資料計算完成分析後續公布資訊
The publisher LAP (Lambert Academic Publishing, Saarbrucken, Germany) was publish the book Logic multicontrollers design. Getting separations of parallel logic control algorithms (in Russian). In comparison with previous publications in this book was exdended description of some algorithms for transformations of parallel graph-schemes and obtaining its numerical estimates, significant number of examples for proposed algorithms was added. A description of some new computational experiments that explore issues of quality of separations and flexibility of parallel-sequential method was given. The last chapter is a description of specialized accelerator for fast processing of R-expressions (a basis of parallel-sequential method), its hardware complexity and computational speed. Description of current computational expreiments carried out under Gerasim@home project does not present in this book — they will be published later after ending of computing using grid and subsequent analysis of the data.

October 16, 2011 感謝所有運算者協助 運算產生了235GB 資料 大約是作者家用需連續運算57年 後續 1~2個月分析data
Dear volunteers! Calculations of the first (and I hope not last) experimental series have been successfully completed, resulting in 235 GB collected raw calculated data. Preliminary postprocessing of this data is completed now (in 3 weeks), first results you can see here. Next, I need 1-2 months for detailed data analysis, publication and taking a final decision about continuation of calculations. Without help of project participants, which combined provide 818 GHz-years of CPU time, I would have need 57 years of continuous operation of my home machine (Core 2 Duo, 1.86 ГГц) to make such calculations (without redundancy). Many thanks to all participants!

May 12, 2012 在某研討會議 發表了 相關報告
Theses of my report "Comparison of methods for the synthesis of separations using two-dimension diagrams" (in Russian) are published at proceedings of conference "Recognition — 2012" (section "Special hardware accelerators, parallel and distributed computing"). This publication focuses at results of qualitative superiority of methods for the synthesis of separations, ie those regions of parameter space where the quality of the one chosen method by the selected criteria is better then analogues.

研討會影像資訊--> http://www.youtube.com/watch?v=9DDi56_P7OY

September 17, 2012 在莫斯科 PACO論壇 初步報告 使用 BOINC 平台進行研究分析 達成了 目前作者的研究成果展現
Report "Using volunteer distributed computing on BOINC platform for quality analysis of separations of parallel logic control algorithms" was adopted by the conference "Parallel computations and control problems (PACO'12)", section "Technology programming parallel and distributed systems". Conference will be held October 24—26 in Institute of Control Sciences Academician V.A. Trapeznikov RAS, Moscow.

November 02, 2012 作者研究 PACO論壇 published
My report was published at proceedings of the conference "Parallel computations and control problems (PACO'12)"

December 25, 2012 發表利用分散式運算進行研究等相關資訊文章
An article "Structural-parametric optimization of logical control systems using distributed volunteer computing" (in Russian) was published!

January 29, 2013 新的研究方向資訊 希望算友(Cruncher)們繼續支持
An article

Vatutin E.I., Titov V.S. Using adjacent neighborhood during greedy sequential separation of parallel graph-schemes (in russian)

is accepted for publication at the journal "Proceedings of the universities. Instrument making". The paper proposes a modification of known S.I. Baranov method used in previous experiments. The modified method is greedy (at the current block included a allowable vertex with minimum value of penalty function) and forms a blocks sequentially (first a first as long as possible, then second and so on). It differs from analog by using restriction for getting vertices only from the neighborhood of the current block. An article includes results of preliminary comparison of methods that obtained from a limited number of selected points at restrictions space. The results show that the proposed method is the best by the number of interconnect links (\alpha) at obtained separations. In addition, during increasing restrictions proposed method shows better than an analog degree of minimizing the number of blocks (H) and intensity of inter-block interaction (\delta) to the detriment of the distribution of microoperation (Y) and logical control signals (X). It is interesting to investigate conservation of this features at all restriction values with any number of vertices in the graph-scheme of algorithm. We decided to conduct two computing experiments

4 <= Xmax <= 150, 1 <= N <= 600
3 <= Wmax <= 200, 1 <= N <= 600

with new method and compare the results with previous results of performed experiments. Hope to support from the crunchers!

依官網 2013/01/31 23:09 頁面資訊 尚有405,272 WU 待分送運算
最後由 alexpon 於 2013-01-31 23:11 編輯,總共編輯了 2 次。
圖檔
圖檔
My State1My State2My State3
My Folding
圖檔
圖檔
圖檔

古詩:手把青秧插滿田 低頭便見水中天 心地清靜方為道 退步原來是向前

家草
學會運用算盤
學會運用算盤
文章: 60
註冊時間: 2006-05-17 00:02
聯繫:

文章 家草 » 2007-07-08 23:22

這個目前是可以加入的

圖檔

BOINC
分散式首腦
分散式首腦
文章: 639
註冊時間: 2006-03-17 23:42
聯繫:

文章 BOINC » 2007-07-11 12:31

目前還沒有確定目標的樣子。
http://www.boinc.ru/forum/viewtopic.php?t=488

回覆文章

誰在線上

正在瀏覽這個版面的使用者:沒有註冊會員 和 1 位訪客