Broken sequences / duplicate entries - Page 2

schickel · 2012-07-03, 08:30

Quote:

Originally Posted by Dubslow

Alright, the following should be a complete list of the 22 sequences with errors that occurred at a size <110. It's likely that there are at least a few sequences with errors at sizes >=110 that I haven't found.

The first three lines are the aliqueit check output, and the fourth is what AllSeq.txt reports is the correct status of the sequence. They're in no particular order, and this includes the others reported above.

Unfortunately, unless you have an older version of the AllSeq file, the status on any broken sequences will be wrong since I just download the elf file and suck the status info off of that. Unfortunately, the errors started turning up before I realized it, so the bad sequence data overwrote the good....

Dubslow · 2012-07-03, 08:34

Quote:

Originally Posted by schickel

Unfortunately, unless you have an older version of the AllSeq file, the status on any broken sequences will be wrong since I just download the elf file and suck the status info off of that. Unfortunately, the errors started turning up before I realized it, so the bad sequence data overwrote the good....

http://www.mersenneforum.org/showpos...&postcount=179, 09 Apr 12, 07:33

Because they were reported to be at roughly the right size, vs. < 100 for all of these errors, I'm pretty sure they're accurate.

In any case, the errors are there, regardless of what the correct sequence is.

schickel · 2012-07-03, 10:03

Problems continue:

.....oops?

Dubslow · 2012-07-03, 22:29

Even though it's broken, I've taken the false downdriver on what isn't 246888 for an absolutely exhilarating ride. The initial run made it down to 45 digits before breaking; it actually caught the 2^2*7 even perfect driver, but amazingly that only lasted 4 lines; 20-30 lines later it recaptured the downdriver. Went down to 25, and then back up to 45, and since then has caught and lost the downdriver another 4-5 times. It's a pretty graph.

The error happened at line 691, a bit after the initial downdriver capture, at about 80 digits. Everything after that isn't actually 246888, but it's fun to watch.

Unfortunately, now it's caught the 2^3*3*5 driver... we'll see if that disappears too. Edit: It did! Wow. But now it's got 2*3. Edit2: Well, 2*3^n for n>0 has taken me back to 85 digits with no end in sight... this roller coaster's over.

PS I think I was wrong about at least one of those 2 sequences I hypothesized to have missed (in post #11). I'll have to wait and see about the other.

Dubslow · 2012-07-08, 00:50

Quote:

Originally Posted by Dubslow

PS I think I was wrong about at least one of those 2 sequences I hypothesized to have missed (in post #11). I'll have to wait and see about the other.

The second one was already on the list, so that list is in fact comprehensive (for errors that occurred at size < 110).

Dubslow · 2012-07-15, 00:55

Okay, it seems Syd did some maintenance today, re-rebuilding the entire Aliquot database. However, this time, my script is acting way funkier than it did during the last rebuild.

Last time it would take about a minute to get each sequence's status, as the DB reconstructed the sequence from the beginning with the numbers already in the DB. This time though, each sequence takes upwards of ten minutes, and after 5-10 sequences, the rest start coming back wrong (i.e., the regex to get basic size and index info doesn't find a match). If I stop and restart the script, I can get another 5-10 seqs (even though they borked out last time) before the next batch starts borking out.

Code:

19:12:43 1 sequences complete: 266466
19:14:07 2 sequences complete: 36468
19:15:35 3 sequences complete: 216468
19:16:05 4 sequences complete: 156468
19:18:25 5 sequences complete: 616470
19:23:16 6 sequences complete: 26472
19:27:27 7 sequences complete: 806472
19:27:37 Seq: 336474 Strange. Termination? More likely a bad sequence.
19:27:37 No basic regex match

Code:

19:39:05 1 sequences complete: 336474
19:47:28 2 sequences complete: 616476
19:51:44 3 sequences complete: 926478

Note how long they all take.

The only reason I'm mentioning this at all is that a few hours ago, the script was behaving just like last time.

Code:

15:11:01 1 sequences complete: 916440
15:13:38 2 sequences complete: 706440
15:14:03 3 sequences complete: 166440
15:14:39 4 sequences complete: 676440
15:15:04 5 sequences complete: 556440
15:15:40 6 sequences complete: 616440
15:16:45 7 sequences complete: 296442
15:17:32 8 sequences complete: 606444
15:18:43 9 sequences complete: 966444
15:20:56 10 sequences complete: 896448
15:21:33 11 sequences complete: 236448
15:22:31 12 sequences complete: 916448
15:23:04 13 sequences complete: 576450
15:23:51 14 sequences complete: 196452
15:27:16 15 sequences complete: 576456
15:27:55 16 sequences complete: 786456
15:28:48 17 sequences complete: 446460
15:29:44 18 sequences complete: 76464
15:30:14 19 sequences complete: 946464
15:31:08 20 sequences complete: 136464

Seqs take < 1 min, as opposed to the > 5 mins I've gotten recently.

fivemack · 2012-07-15, 01:23

Ah: I ran 434040 to 125 digits, and I fear I may have run it from a wrong start position. I started from term 1 a couple of days ago and am up into the term number 1600s by now, but I don't know whether I'll merge into the bit I calculated earlier.

Batalov · 2012-07-15, 01:43

It should be

Code:

434040 125      1739 .  147953024.. = 2^4 * 7 * 99523 * 499071109.. * 266454332.. * ...

(I am running some scripts as well)

Btw, even in the incomplete pull (it was interrupted in 600s with the DB's "Offline for a few minutes. Resetting sequence cache."), there were more than those 22 in the "comprehensive" list

Dubslow · 2012-07-15, 01:56

Quote:

Originally Posted by Batalov

It should be

Code:

434040 125      1739 .  147953024.. = 2^4 * 7 * 99523 * 499071109.. * 266454332.. * ...

(I am running some scripts as well)

Btw, even in the incomplete pull (it was interrupted in 600s with the DB's "Offline for a few minutes. Resetting sequence cache."), there were more than those 22 in the "comprehensive" list

Those are the sequences that appeared as size < 110 when I initially filled my webpage. All the rest were at 111+ digits. I said it was comprehensive for errors that occurred at smaller sizes, and I hypothesized that there were others with errors above size 110 that I missed.

In any case, fivemack's 434040 is in the list I gave.

Batalov · 2012-07-15, 02:01

Quote:

Originally Posted by Dubslow

Those are the sequences that appeared as size < 110 when I initially filled my webpage. All the rest were at 111+ digits. I said it was comprehensive for errors that occurred at smaller sizes, and I hypothesized that there were others with errors above size 110 that I missed.

That hypothesis didn't hold water. There were broken sequences where the broken tail was well above 110 digits, yet they were broken at sizes as small as 20 digits. To find broken sequences you have to run them - not just look at the size.

Checking a sequence is ~100 times faster than retrieving it!

Dubslow · 2012-07-15, 02:05

Quote:

Originally Posted by Batalov

That hypothesis didn't hold water. There were broken sequences where the broken tail was well above 110 digits, yet they were broken at sizes as small as 20 digits. To find broken sequences you have to run them - not just look at the size.

Yes, but given 9207 sequences to choose from, I wasn't exactly about to go searching for those sorts of errors. (I also assumed that any tail would be no more than a few lines long, because those numbers wouldn't be in the database, but obviously I forgot that some errors are old enough that workers extended the errors without realizing it.)

2012-07-03, 10:03	#14
schickel "Frank <^>" Dec 2004 CDP Janesville 84A₁₆ Posts	Problems continue: .....oops? Attached Thumbnails

2012-07-03, 22:29	#15
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	Even though it's broken, I've taken the false downdriver on what isn't 246888 for an absolutely exhilarating ride. The initial run made it down to 45 digits before breaking; it actually caught the 2^27 even perfect driver, but amazingly that only lasted 4 lines; 20-30 lines later it recaptured the downdriver. Went down to 25, and then back up to 45, and since then has caught and lost the downdriver another 4-5 times. It's a pretty graph. The error happened at line 691, a bit after the initial downdriver capture, at about 80 digits. Everything after that isn't actually 246888, but it's fun to watch. Unfortunately, now it's caught the 2^335 driver... we'll see if that disappears too. Edit: It did! Wow. But now it's got 23. Edit2: Well, 23^n for n>0 has taken me back to 85 digits with no end in sight... this roller coaster's over. PS I think I was wrong about at least one of those 2 sequences I hypothesized to have missed (in post #11). I'll have to wait and see about the other. Last fiddled with by Dubslow on 2012-07-03 at 22:55*

2012-07-15, 01:43	#19
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 2×7×677 Posts	It should be Code: 434040 125 1739 . 147953024.. = 2^4 * 7 * 99523 * 499071109.. * 266454332.. * ... (I am running some scripts as well) Btw, even in the incomplete pull (it was interrupted in 600s with the DB's "Offline for a few minutes. Resetting sequence cache."), there were more than those 22 in the "comprehensive" list Last fiddled with by Batalov on 2012-07-15 at 01:47

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Broken aliquot sequences	fivemack	FactorDB	46	2021-02-21 10:46
Graphs broken for aliquot sequences on factordb	fivemack	Aliquot Sequences	4	2018-03-23 11:54
bugreport: duplicate entries	maxal	FactorDB	1	2017-11-14 15:38
Broken aliquot sequences	schickel	FactorDB	18	2013-06-12 16:09
Wrong entries in database?	Unregistered	Information & Answers	4	2008-06-28 01:50

2012-07-15, 01:23	#18
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 7²·131 Posts	Ah: I ran 434040 to 125 digits, and I fear I may have run it from a wrong start position. I started from term 1 a couple of days ago and am up into the term number 1600s by now, but I don't know whether I'll merge into the bit I calculated earlier.