Second valley may be lower

[Kileran]

I've seen a few people commenting lately that they feel "gipped" by having to wait through all 250 gens when they hit thier lowest valley at gen 71. Many people have commented that there is a possibility that you will reach yet another valley before the end.

I just wanted to add my support for letting it run. i was in the same situation. i've been sitting here for 4 days thinking "why am i waiting when i could start a better one"

I just got home, i'm at gen 221, and it's probably going to drop to a new low. i havn't seen a low like this since 113. I'm glad i didn't ditch out early.

Kileran

[FoBoT]

uh, if you look at the top ten best structures, 6 of them are gen. 250

i think that says a lot for waiting for a complete run to finish

[m0ti]

the top 10 shows what generation the person is currently on within the batch for which the best fold was found.


So it first appears at gen X when they find the fold. But it increases up to gen 250 as the person completes all the generations.

[Mikus]

Just took a look at the 'View Details' of the current top 10. Considering ONLY the 'RMSD vs. Time' graphs :

Two of the ten had their best RMSD beyond generation 200.

NONE of the ten appeared to be descending into a second valley. [Maybe if a cycle were 500 generations long instead of 250 generations ?]

As I wished in another thread, the '# of generations' field in the top-10 table OUGHT to show the generation at which the RMSD value (as shown in that line) was reached.

mikus

[FoBoT]

Originally posted by m0ti
the top 10 shows what generation the person is currently on within the batch for which the best fold was found.


So it first appears at gen X when they find the fold. But it increases up to gen 250 as the person completes all the generations.

oh, nevermind :bonk:

[Grumpy]

I have found that at the last 20-30 Gens it will sometimes suddenly drop markedly..sadly for my poor RMSD, not markedly enough :cry:

[gistech1978]

i think the science is more important than stats
i like stats as much as the next guy, but thats not why i do DC.

[tpdooley]

The stats freaks have some valid points - they want it to run fast, not require babysitting, and to cut off unneccessary generations that don't provide statistically useful data.

[gistech1978]

good points tpdooley that i didnt take into consideration.

[Brian the Fist]

We do not know if the algorithm will 'revisit' second minima until we try it, remember the algoritm is still subject to minor modifications. We will be tweaking a few parameters over the next little while to learn what works best.

[prokaryote]

Hi Dr. Feldman,

Any thoughts on This (http://www.free-dc.org/forum/showthread.php?s=&threadid=3450) thread? Just curious.

Thanks,

prok

[prokaryote]

bump