Validation inconclusive?


log in

Advanced search

Message boards : Number crunching : Validation inconclusive?

Author Message
Profile Kibble (KB7TIB)
Avatar
Send message
Joined: 6 Dec 99
Posts: 21
Credit: 1,921,543
RAC: 4,672
United States
Message 927737 - Posted: 21 Aug 2009, 16:10:07 UTC

I'm not sure what the result of a pending WU means when two results are returned, from separate sources, of course. Does this mean, perhaps, that there is something wrong with the validation process. This is the first time I've ever seen this notation. http://setiathome.berkeley.edu/workunit.php?wuid=492785162


____________

Profile [B^S] madmac
Volunteer tester
Avatar
Send message
Joined: 9 Feb 04
Posts: 1158
Credit: 3,956,225
RAC: 3,115
United Kingdom
Message 927748 - Posted: 21 Aug 2009, 16:28:14 UTC
Last modified: 21 Aug 2009, 16:28:50 UTC

No it does not mean that there is something wrong with validation process, the results that you and your wingman has done or not the same. I know because I am waiting for one, the results are close but not exactly the same. The WU is then posted to a third and when that comes back you will get the credit so I am afraid that it could be another month before you get credit or be unlucky and wait longer.
____________

Profile dnolanProject donor
Avatar
Send message
Joined: 30 Aug 01
Posts: 1255
Credit: 46,514,138
RAC: 2,867
United States
Message 927771 - Posted: 21 Aug 2009, 17:02:21 UTC - in response to Message 927737.

I'm not sure what the result of a pending WU means when two results are returned, from separate sources, of course. Does this mean, perhaps, that there is something wrong with the validation process. This is the first time I've ever seen this notation. Link to WU


(made link active)
To have a pending result with 2 or more results returned is not that unusual.
The situation can mean one of several things:
- You had an error in your calculation, you won't get credit.
- The other person had an error, that person won't get credit.
- Both of you had errors, neither one of you will get credit.
- Neither one of you made a mistake, but the results aren't strongly similar, either or both will eventually get credit, depending on if the two results already in are weakly similar or not.

The amount of time you have to wait to find out depends on how long it takes to get 2 results returned that are strongly similar, once that happens, you'll find out which of the above cases are true.

-Dave
____________

Ianab
Volunteer tester
Send message
Joined: 11 Jun 08
Posts: 678
Credit: 12,769,447
RAC: 1,939
New Zealand
Message 927831 - Posted: 21 Aug 2009, 21:59:59 UTC

When you look at the actual results returned you see.

Your machine:

Spike count: 15
Pulse count: 6
Triplet count: 8
Gaussian count: 0

Wingman

Spike count: 0
Pulse count: 0
Triplet count: 0
Gaussian count: 0

Very different results, so the validation IS working as the results didn't match. I would suspect yours is correct and the wingman had an error.

Another machine has errored out, and it's now been sent to a 4th machine. If that result agress with yours, then you get the credits.

Ian

Profile Kibble (KB7TIB)
Avatar
Send message
Joined: 6 Dec 99
Posts: 21
Credit: 1,921,543
RAC: 4,672
United States
Message 927888 - Posted: 22 Aug 2009, 1:47:48 UTC - in response to Message 927831.

Thank you all; I understand the problem now. I was only surprised since this was the first time I've ever seen that particular result. Hopefully the next result will clear quickly for SETI@home's sake.
____________

Josef W. SegurProject donor
Volunteer developer
Volunteer tester
Send message
Joined: 30 Oct 99
Posts: 4347
Credit: 1,124,780
RAC: 786
United States
Message 927890 - Posted: 22 Aug 2009, 1:49:13 UTC - in response to Message 927831.

When you look at the actual results returned you see.

Your machine:

Spike count: 15
Pulse count: 6
Triplet count: 8
Gaussian count: 0

Wingman

Spike count: 0
Pulse count: 0
Triplet count: 0
Gaussian count: 0

Very different results, so the validation IS working as the results didn't match. I would suspect yours is correct and the wingman had an error.

No, the wingman is using optimized "SSE3 Win32 Build 41", and like all AK_v8 builds before rev. 52 it doesn't preserve individual signal counts in the checkpoint file. Thus the counts are only those after its "Restarted at 70.57 percent". It's result file may well have 29 signals too, with one just enough different that a "strongly similar" comparison wasn't achieved.

Another machine has errored out, and it's now been sent to a 4th machine. If that result agress with yours, then you get the credits.

Ian

There's no way yet to judge how it'll turn out. I think most of these cases end up with 3 hosts getting credit, but don't have enough data to even guess at the statistical probabilities.
Joe

Ianab
Volunteer tester
Send message
Joined: 11 Jun 08
Posts: 678
Credit: 12,769,447
RAC: 1,939
New Zealand
Message 927942 - Posted: 22 Aug 2009, 8:02:22 UTC - in response to Message 927890.

Thanks for clearing that up Joe.

Hopefully it's just a slightly different result from the different versions...

Ian

Profile LiliKrist
Volunteer tester
Avatar
Send message
Joined: 12 Aug 09
Posts: 333
Credit: 143,167
RAC: 0
Indonesia
Message 928138 - Posted: 23 Aug 2009, 4:08:25 UTC

I have the same matter.

Here my task result :

Task ID : 1337788497
WU ID : 493423957
Sent : 18 Aug 2009 13:25:05 UTC
Time Reporded : 19 Aug 2009 14:53:40 UTC
Status : Completed, validation inconclusive
Run Times : 38,374.75
Claimed Credit : 85.93
Granted Credit : pending

Profile burnz
Avatar
Send message
Joined: 14 Apr 04
Posts: 26
Credit: 178,564
RAC: 0
Australia
Message 928162 - Posted: 23 Aug 2009, 8:42:41 UTC

three inconclusive's i have not crunched it yet.
http://setiathome.berkeley.edu/workunit.php?wuid=494998204
so they are waiting on me.
____________

1mp0£173
Volunteer tester
Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 928200 - Posted: 23 Aug 2009, 16:17:30 UTC - in response to Message 927942.

Thanks for clearing that up Joe.

Hopefully it's just a slightly different result from the different versions...

Ian

It is far more likely that it is different results from slightly different processors.

Most of the floating point math done by modern CPUs (FPUs) are estimated. The estimates are very, very good, but they're estimates. If one FPU (even from the same vendor, but different processor versions) consistently estimates just a tiny bit high, and the other estimates just a tiny bit too low, those errors can accumulate just enough to "fail" validation.

Give the validators another result (or two) and they'll all center up around some actual value and eventually validate (probably).

If it happens alot, it could be a problem with your CPU, but one robin does not make it spring.
____________

Message boards : Number crunching : Validation inconclusive?

Copyright © 2014 University of California