Timeline for Examples of errors in computational combinatorics results
Current License: CC BY-SA 4.0
21 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 19, 2023 at 23:22 | answer | added | Timothy Chow | timeline score: 2 | |
| Jan 17, 2023 at 7:04 | history | edited | Martin Sleziak | CC BY-SA 4.0 |
http -> https (the question has been bumped anyway)
|
| Jan 16, 2023 at 16:06 | answer | added | The Amplitwist | timeline score: 1 | |
| Jan 13, 2023 at 7:42 | comment | added | Jukka Kohonen | @Jean Marie Becker: Sorry for being unclear, that's just MathOverflow slang for "community wiki", basically meaning that votes do not affect reputation (and some other details). Big-list questions are typically made such (it seems). | |
| Jan 13, 2023 at 6:13 | comment | added | Jean Marie Becker | Could you tell me the meaning of "being CW". I don't know the meaning of this abbreviation (not native English speaker) ? | |
| Jan 13, 2023 at 2:34 | comment | added | Timothy Chow | Again this is not exactly computational combinatorics but rather enumerative geometry. Famously, the calculation of the number of twisted cubics on a generic quintic hypersurface was done in two different ways, one via "physics methods" that were (at the time) nonrigorous, and one via conventional mathematical methods. Initially the two answers disagreed. It turned out that the "rigorous" mathematical calculation was wrong because of a bug in the program. | |
| Jan 13, 2023 at 2:23 | comment | added | Timothy Chow | Regarding how to avoid errors, a simple technique is to do two independent computations of the same value. This technique seems to have worked well for GIMPS, for example, and for computing the digits of $\pi$. | |
| Jan 13, 2023 at 2:12 | comment | added | Timothy Chow | Perhaps not combinatorics per se, but see Computing $\pi(x)$: the Meissel-Lehmer method, by Lagarias, Odlyzko, and Miller. It was a running joke that the largest value of every published table of values of $\pi(x)$ would be wrong, until Lagarias, Odlyzko, and Miller "broke the curse." | |
| Jan 13, 2023 at 1:15 | answer | added | BPP | timeline score: 15 | |
| Jan 12, 2023 at 14:30 | answer | added | Ben Burns | timeline score: 4 | |
| Jan 12, 2023 at 11:27 | answer | added | Olaf Teschke | timeline score: 7 | |
| Jan 12, 2023 at 9:09 | history | edited | Jukka Kohonen | CC BY-SA 4.0 |
extending to allow reputable arXiv manuscripts & explain why excluding OEIS-only changes
|
| Jan 12, 2023 at 8:11 | answer | added | Christian Stump | timeline score: 7 | |
| Jan 11, 2023 at 21:21 | answer | added | colt_browning | timeline score: 9 | |
| Jan 11, 2023 at 19:31 | history | became hot network question | |||
| Jan 11, 2023 at 14:09 | answer | added | Brendan McKay | timeline score: 21 | |
| Jan 11, 2023 at 13:36 | comment | added | Jukka Kohonen | @Peter: I would be interested in incorrect bounds as well! (But not about correct claims that have simply been improved later...) The key thing I am looking for is dependence on substantial computation, and the following opacity of the published claims, as in "you have to believe us, because we computed it". | |
| Jan 11, 2023 at 12:50 | comment | added | Peter Taylor | What about incorrect claims of bounds? The specific example I'm thinking of fails 5 because the correction was only published in OEIS, but the published paper contained the first few terms of a sequence and "the next term being larger than [Y]" where the next term was in fact considerably smaller than Y. | |
| Jan 11, 2023 at 12:40 | history | made wiki | Post Made Community Wiki by Stefan Kohl♦ | ||
| Jan 11, 2023 at 11:47 | history | edited | Jukka Kohonen | CC BY-SA 4.0 |
One more MO question reference
|
| Jan 11, 2023 at 11:26 | history | asked | Jukka Kohonen | CC BY-SA 4.0 |