 |
| Bluumi [SwissTeam.NET] 2002-09-08 15:21:28 | Hello..... With a dupeChecker from Banenenweizen, i saw thing i could not accept.... Why can Muon with the SAME Parameters make OTHER results?? Look in the official TOP250.... - results 58 and 89 - results 120 and 128 - results 14 and 138 - results 151 and 182 - results 151 and 183 and the best!! - results 12 and 218 1.718980 (47129 particles) [v4.21b] {B75C17BC} Rank 216 by [SG]_DOA has the SAME parameters (all !) as 1.856586 (47206 particles) [v4.21b] {B75C17BC} Rank 12 by [SG]proteino.de its CRAZY ... sorry... but can this be? And if only there are SO many dupes... how many dupes with DIFFERENT results can it do? What can made a 1.86 from a 1.72 result? Heat? ShowWave? PsyWar?   [This message was edited by Bluumi [SwissTeam.NET] on 2002-Sep-08 at 22:40.] |
| David Bass 2002-09-08 15:49:20 | Is this not related to expected random variations in the decay of the pions into muons? These, AIUI are not pre-programmed, but occur according to the predicted half-life, and with random variations in the direction of the output muon. Stephen seems to do something clever to minimise the differences between runs by using weighting, but variations will still occur. In short, not a bug but expected statistical variation. It might be wise to cull such duplicates from the top 250, to improve the breadth of the gene pool (as it were). On a related note - the variations in scores for these duplicate configurations provide a further hint that the latest (2.9+%) figure is likely to be signficantly different to Stephen's monster result. |
| Stephen Brooks 2002-09-09 01:53:04 | The variation in results using the same design is about ±0.1% here generally. In the last version of the program it was ±0.3%, which was unacceptably big. Also it appears that the extent of the variation changes with the design, so the "extreme" design I put in only seems to have a characteristic variation of ±0.02% or something because maybe the larger solenoids make it less easy for changes in the pion decay to make a difference to the number getting through. The decays produce a "course change" to the particle, which would affect the results much more greatly when the solenoids are small so the particles are aiming at a "small target". Large solenoids all the way down have a large enough field region to bend a lot of particles back in however they decay. "As every 11-year-old kid knows, if you concentrate enough Van-der-Graff generators and expensive special effects in one place, you create a spiral space-time whirly thing, AND an interesting plotline" |
| Bluumi [SwissTeam.NET] 2002-09-09 03:57:50 | What, if "bad" parameters give a good %-Result, and a "super" parameter a bad %Result.... The evolution code will help the bad-Parameters because the % are better? OR does it try both, the best and some "bad"-% Results ? Can it go a wrong way, if a bad Parameter Field is chosen? cya |
| Stephen Brooks 2002-09-09 07:10:20 | There is a limit to how far the percentages can vary. A 2.0% design might score 1.9% or possibly 1.8%, but it definitely won't be beaten by a design that is meant to score, say, 1.25%. The algorithm does not just always pick the _best_ results (also a few lower ones), so other good parameters that score unfairly low will probably get tried again. |
| Bluumi [SwissTeam.NET] 2002-09-09 12:49:44 | Ok, accept ... i will give my power again  |
