Error Rates

Prime95 · 2002-08-19, 19:39

Now that the list of bad LL test results has been posted, would anyone care to analyze the error rates? Perhaps broken down by program version and whether errors were reported during the run. Also, is the error rate increasing with exponent size?

Such a volunteer will need to week out all the duplicates and probably analyze only exponents below 5.2 million (the current limiit for exponents with matching LL tests).

This anaylsis would help answer the question "What is the probability of a Mersenne prime being discovered out of order?"

Kevin · 2002-08-19, 21:46

Just so people don't have to search, where is the list of bad LL results posted?
nm, i think i found it
ftp://mersenne.org/gimps/bad.zip

binarydigits · 2002-08-19, 22:43

The probability of a Mersenne prime being discovered out of order is 100% because it's already happened :~)
(Sorry, couldn't resist - I know you meant "another one".)

That analysis sounds like something right up my alley, but I'm leaving on vacation tomorrow so I couldn't even start on it until September.

I assume you have already tossed any results from that bad version from about three years ago?

Prime95 · 2002-08-19, 23:07

Quote:

Originally Posted by binarydigits

I assume you have already tossed any results from that bad version from about three years ago?

Yes, those are in a different file (so users still receive CPU credit). This is the one exception to the rule of bad results get no CPU credit in my rankings.

outlnder · 2002-08-19, 23:41

Let me try to wrap my pea brain around this, ALL those exponents that are "bad" have not been proven to be non- prime???

If so, why aren't we doing those as fast as possible now??

I would be glad to volunteer to do these exponents for you.

Prime95 · 2002-08-19, 23:46

Quote:

Originally Posted by outlnder

Let me try to wrap my pea brain around this, ALL those exponents that are "bad" have not been proven to be non- prime???

If so, why aren't we doing those as fast as possible now??

The file bad.zip has LL results that were proven wrong by a triple-check.
The lucas_v.zip file has the two correct results.

Thus, you can calculate prime95's error rate by analyzing a range of exponents (say 3 million to 4 million) and count the number of unique entries in bad.zip and divide that by the entries in lucas_v and bad.

That is, error_rate = bad / (bad + good)

outlnder · 2002-08-19, 23:55

OK, I get it now.

On another note, TPR has some exponents that need triple checking. This looks like a negative for the team. Is it possible to ask to do the triple check so as to remove this blotch from our record???

ops:

garo · 2002-08-20, 01:41

I can do some analysis on that. I was looking at the bad.zip file for some trends the other day so I can continue on the job.

outlnder,
That question really belongs in the Ars forum. Anyway that list has not been updated for some time. Rob needs to work on it since I know that my exponent on that list was verified and "proven" in my favor.
Garo.

outlnder · 2002-08-20, 02:41

I believe this to be the correct Forum since I am asking George.

Isn't it up to George whether or not I can request certain exponents??

Digital Concepts · 2002-08-20, 02:49

To ensure objectivity, triple checks should go to a third party.

Kevin · 2002-08-20, 02:58

Quote:

Originally Posted by outlnder

Isn't it up to George whether or not I can request certain exponents??

There are ways to do it by yourself, without permission, but it's still probably better to ask and make sure it's done right.

2002-08-19, 19:39	#1
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 2×53×71 Posts	Error Rates Now that the list of bad LL test results has been posted, would anyone care to analyze the error rates? Perhaps broken down by program version and whether errors were reported during the run. Also, is the error rate increasing with exponent size? Such a volunteer will need to week out all the duplicates and probably analyze only exponents below 5.2 million (the current limiit for exponents with matching LL tests). This anaylsis would help answer the question "What is the probability of a Mersenne prime being discovered out of order?"

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Electrical Service Rates	storm5510	Hardware	70	2019-10-14 21:14
error rates and P-1 test	drakkar67	Prime Sierpinski Project	9	2008-05-26 14:29
error rates	drakkar67	Prime Sierpinski Project	12	2006-04-21 17:26
LL Test Rates and GIMPS Promotion	Primenut	Lounge	14	2003-06-09 09:32
Error rates revealed	Prime95	Math	1	2002-09-01 00:10

2002-08-19, 21:46	#2
Kevin Aug 2002 Ann Arbor, MI 661₈ Posts	Just so people don't have to search, where is the list of bad LL results posted? nm, i think i found it ftp://mersenne.org/gimps/bad.zip

2002-08-19, 22:43	#3
binarydigits Aug 2002 64₈ Posts	The probability of a Mersenne prime being discovered out of order is 100% because it's already happened :~) (Sorry, couldn't resist - I know you meant "another one".) That analysis sounds like something right up my alley, but I'm leaving on vacation tomorrow so I couldn't even start on it until September. I assume you have already tossed any results from that bad version from about three years ago?

2002-08-19, 23:41	#5
outlnder Aug 2002 100111110₂ Posts	Let me try to wrap my pea brain around this, ALL those exponents that are "bad" have not been proven to be non- prime??? If so, why aren't we doing those as fast as possible now?? I would be glad to volunteer to do these exponents for you.

2002-08-19, 23:55	#7
outlnder Aug 2002 2×3×53 Posts	OK, I get it now. On another note, TPR has some exponents that need triple checking. This looks like a negative for the team. Is it possible to ask to do the triple check so as to remove this blotch from our record??? ops:

2002-08-20, 01:41	#8
garo Aug 2002 Termonfeckin, IE 2²·691 Posts	I can do some analysis on that. I was looking at the bad.zip file for some trends the other day so I can continue on the job. outlnder, That question really belongs in the Ars forum. Anyway that list has not been updated for some time. Rob needs to work on it since I know that my exponent on that list was verified and "proven" in my favor. Garo.

2002-08-20, 02:41	#9
outlnder Aug 2002 2×3×53 Posts	I believe this to be the correct Forum since I am asking George. Isn't it up to George whether or not I can request certain exponents??

2002-08-20, 02:49	#10
Digital Concepts Aug 2002 110110₂ Posts	To ensure objectivity, triple checks should go to a third party.