Anonymous user
Talk:Percolation/Bond percolation: Difference between revisions
→D Entry
| (3 intermediate revisions by 3 users not shown) | |||
Line 15:
: I added the error note, but somehow the reason part of the template doesn't show. It's this: at p=.5, the result should be somewhere around .5 too, but the D code gives 0.14. Given that it's the average of 10000 runs, it's impossible that that was a random fluctuation, so there's got be a bug in the code. --[[User:Ledrug|Ledrug]] ([[User talk:Ledrug|talk]]) 03:39, 6 March 2014 (UTC)
:: Thank you Ledrug, I'll try to fix that. -[[User:Bearophile|bearophile]] ([[User talk:Bearophile|talk]])
:: The D entry is a translation of the Python entry, with some refinements. It seems the Python entry gives the same wrong results: http://ideone.com/ik57yE And is the C entry having the same problems?
::: You lost me. All entries other than D gives about 0.5 at p = 0.5, including the ideone output. Given 1000 tries, that's about 500 successes on average, so random fluctuation sigma is sqrt(500) ~ 23. At 3 sigma, i.e. 99.7% confidence, a correct solution should show between roughly 430 to 570 pass rate. D shows 142 out of 1000, that's 15 sigmas away, which you need lottery-winning kind of luck to get.
::: It doesn't matter how the D code comes about; it shows result different from everyone else's and which is extremely statistically implausible, so the natural conclusion, at near 100% confidence even, is there's a bug. --[[User:Ledrug|Ledrug]] ([[User talk:Ledrug|talk]]) 22:33, 9 March 2014 (UTC)
:::Hi Bearophile, you do state in the D entry that you make performance optimisations over the Python. Maybe the problem is there? --[[User:Paddy3118|Paddy3118]] ([[User talk:Paddy3118|talk]]) 23:09, 9 March 2014 (UTC)
:::: Rewritten D entry.-[[User:Bearophile|bearophile]] ([[User talk:Bearophile|talk]])
| |||