| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Nutritious Rice for the World
Thread: Minimum quorum reduced from 14 to 10 days for the Rice project |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 9
|
|
| Author |
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
Since June 24, 2008 the minimum quorum for getting a Rice WU validated has been reduced from 14 to 10.
----------------------------------------This is the most visible of several measures currently taken by the techs to improve the throughput for this project and to get a more homogeneous level of details fot the results regardless of the difficulty of the genes involved. As a side effect results should be validated sooner since the initial distribution is still 19. |
||
|
|
Trong Nguyen
Cruncher Joined: Jun 26, 2008 Post Count: 2 Status: Offline |
Is there a chance that in the future, the initial distribution will be reduce also? I crunch RICE exclusively for a while and notice that in most cases, after a result get validated, there're still 4 or 5 copies of that task still "In progress".
 |
|||
|
||||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
There's always looking at what the best minimum is without compromising scientific value. What i understand is that the results are monitored for sufficient structure, thus yes, there is chance. All depends on e.g. the average mix of computers that return results. It's a very very light project, so you really don't want to dump these over onto super duper computers i.e. leave enough also for the weaker shoulders to have their part of the Rice project.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
Is there a chance that in the future, the initial distribution will be reduce also? I crunch RICE exclusively for a while and notice that in most cases, after a result get validated, there're still 4 or 5 copies of that task still "In progress". Please remember that every result returned for Rice is unique. With Rice we do not collect the results for the workunit until all results have reached a end state (late, error, success). Then ALL the valid results are sent back to the researchers for them to analyze (regardless if they were returned before validation was run the first time or not). We do some statistics on the rate of results being returned and we expect that a majority portion of those results still 'in progress' even after validation will be returned so that we can collect and return to the scientists. They want about 120,000 predictions per gene. Each result contains between 50 and 500 structures depending on the difficulty of the gene and power of the computer. The way we do the analysis we are able to keep the structure generated pretty close to the desired target. Setting the quorum to 10 simply means that you get your points quicker and it slightly changes our stats computation. [Edit 1 times, last edit by knreed at Jul 30, 2008 7:16:15 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
knreed,
I'm somewhat confused... as I understand it, several copies of a Rice WU are sent out for computational analysis. So let's say, 10 machines get it dispatched to them. Now, each of these machines have differing computing capacity because their cpu speeds; memory speeds; quantity of memory; quantity of HDD work space are different. Where I get lost in this is these WU's are rigged to run for approximately 8.00 hrs +/- about 0.10 units of time. So, if there 10 machines that are increasingly more powerful than the last, how does the most powerful computers results ever get completely validated? Is it just a "well, all the other data from the machines below it match everything the fastest machine provided, hence, the data the fastest machine provided must be correct kind of syndrome?" TIA |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Barney, every copy is different. None of the work units do the same work.
Validation is done without comparing for an exact match. Please read the project FAQ if you want to know more. |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges:
|
Barney,
----------------------------------------Read through the info on these two pages (in particular the bottom section of the last page) http://www.worldcommunitygrid.org/projects_showcase/rice/viewRiceMain.do http://www.worldcommunitygrid.org/projects_showcase/rice/viewRiceAbout.do The researchers take the structures produced and do some posting processing on the structures. Essentially, they are looking for clusters of predicted structures. Incorrectly computed structures show up and noise and are thus filtered out and therefore they can handle some error from the results. We actually compute more than the 100,000 that they ask for to ensure that they have enough good results. We also added some error checking to the code so we can catch a large portion of the incorrect results. -------- It is worth noting that a random seed is used by the application to initialize some starting parameters for each structure prediction. This is what makes each result unique. [Edit 1 times, last edit by knreed at Jul 30, 2008 8:30:42 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Barney, if you are interested in this topic, try reading about Monte Carlo simulation.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
krneed & Didactylos ,
thanks I appreciate your responses. Yes, I've jacked with Monte Carlo algorithms in the past.. been a few years but nevertheless, had to do it for one reason or another. Just wasn't completely sure how this was being done. All I could see was that differing amounts of data per "WU" in a rice WU would likely be presented, presuming the initial dispatch would be the same. But there's always a new twist. Thanks again for the responses. |
||
|
|
|