I have been wondering something. Is the same WU downloaded to more than one person? Since there is no certainty that if you only run 10 or 15 models that you will get the lowest possible energy for a protein, this would seem like the most logical way. If several people run the same WU hundreds of models could be generated.

---

PUDDIN TAME

Hi Jim,

every WU is sent out numerous times. If you go here, you find your results. In teh column '#Predictions user' you find the number of decoys you have computed, in '#Predictions all users' you see, how many decoys have been computed collectively.

Regards,

Christoph

Note that each user will have a slightly different workunit, although it's based on the same protein structure and method, it will have a differnet random seed, so each calculation will be done with a different "path" across the landscape. That means that each calculation done is unique (or at least, when I've done 800 units out of 350000 calculated for one protein, it's not likely to be the same as many others).

--
Mats

---

Hi Jim,

every WU is sent out numerous times. If you go here, you find your results. In teh column '#Predictions user' you find the number of decoys you have computed, in '#Predictions all users' you see, how many decoys have been computed collectively.

Regards,

Christoph

Sorry to be pedantic Christoph. You have the right idea but are using the words in a slightly non-standard way, along with a lot of people on this project.

Every WU type is sent out numerous times.

Two tasks from the same WU type have the same data files and the same stategy for finding protein structures, but will have different random seeds.

On this project a WU is usually sent out only once, called a task. If that task comes back with an error (or user abort) then a second task is sent out from that WU. In that case the second task wil have the same seed as the first.

The main reason it matters to get the terminology right is for users who (now or later) choose to run other BOINC projects as well.

On other projects a WU is sent out as a number of identical tasks which should return the same result (within rounding errors). As explained above, this project chooses not to do that, as it has other ways of excluding science errors and prefers to get twice as many samples done instead of double crunching each.

River~~

---

Hi Jim,

every WU is sent out numerous times. If you go here, you find your results. In teh column '#Predictions user' you find the number of decoys you have computed, in '#Predictions all users' you see, how many decoys have been computed collectively.

Regards,

Christoph

Thanks for the Info.

jim

---

PUDDIN TAME

Note that each user will have a slightly different workunit, although it's based on the same protein structure and method, it will have a differnet random seed, so each calculation will be done with a different "path" across the landscape. That means that each calculation done is unique (or at least, when I've done 800 units out of 350000 calculated for one protein, it's not likely to be the same as many others).

--
Mats

Thanks for the Info

Jim

---

PUDDIN TAME

Hi Jim,

every WU is sent out numerous times. If you go here, you find your results. In teh column '#Predictions user' you find the number of decoys you have computed, in '#Predictions all users' you see, how many decoys have been computed collectively.

Regards,

Christoph

Sorry to be pedantic Christoph. You have the right idea but are using the words in a slightly non-standard way, along with a lot of people on this project.

Every WU type is sent out numerous times.

Two tasks from the same WU type have the same data files and the same stategy for finding protein structures, but will have different random seeds.

On this project a WU is usually sent out only once, called a task. If that task comes back with an error (or user abort) then a second task is sent out from that WU. In that case the second task wil have the same seed as the first.

The main reason it matters to get the terminology right is for users who (now or later) choose to run other BOINC projects as well.

On other projects a WU is sent out as a number of identical tasks which should return the same result (within rounding errors). As explained above, this project chooses not to do that, as it has other ways of excluding science errors and prefers to get twice as many samples done instead of double crunching each.

River~~

Thanks for the Info

Jim

---

PUDDIN TAME