Is a Quorum of 1/4 really necessary?

Quorum = 1

4 are initiated, but my understanding is they\'re not identical, and the result is not matched. The \'initial condition parameter\' in the list of model parameters introduces a random number into the models.

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Hi Rebirther

If you want us to look at the details of a particular workunit you\'re crunching, you\'ll need to let us see your computer(s)!

---

Cpdn news

Hi Rebirther

If you want us to look at the details of a particular workunit you\'re crunching, you\'ll need to let us see your computer(s)!

Oh sorry, was a long time to do something here :o

As I understand it, the replication of 4 means that 4 copies with the same model parameters were made. All four were distributed within about a week but John G**** and Anonymous haven\'t trickled yet, so we don\'t know what will happen there. Dr**y and you are trickling.

I think Oxford probably want to compare the results of this model on more than one computer. To be more sure that they will be able to compare 2 results, they need to send the model to more than 2 computers because on average only about 25% of models are completed.

Maybe your model is more important than some other models. Very often only one copy is sent out; another copy is only sent out to a different computer if the first copy crashes or doesn\'t trickle for 6 weeks.

I think the quorum means the minimum number of this set of 4 that must be sent out.

---

Cpdn news

The project people here often use \"data slots\" for a purpose different to that intended, mostly because there isn\'t anywhere for placing the info they DO wish to use. So the \"quorum\" slot was just a convenient place to record the details of the different copies of a model.
As Mo said, this is because they really want data for that parameter set, so they send it out several times. Why it\'s important varies from time to time.

As for the original intended use of \"quorum\", in this project it\'s every man & model for themselves.
\"Your\" results may well get compared to \"others\" based on the same dataset, but only to see what happens to it on different brands of cpu / different versions of cpu / etc / etc. (Which may have different maths routines / different internal accuracies / etc.)