Profile
Best ForComments
Mon, 01/16/2012 - 03:41
#2
shake/wiggle loop does not converge?
The reason I was looking into this is that earlier today I let this script run dozens of iterations against a fold of the Rv0657c contest. Once I checked the "show script commands" box in the "show script output" window I was certain it was looping in the bottom half of fsl.MakeStable. I was expecting the loop to converge, but even *within the loop* the fold would randomly walk between (as I recall) 9430 and 9460 points (a range of 30 points), and the sequence of deltas would be something like 0.8, 11, 2, 5, 0.9, 23, 2, 6, 9, 18, 0.1, 3 ... Obviously some of the deltas were negative and some positive, but the script output does not tell you which were which. I know the script doesn't reset the clashing importance and it is possible I might have run it after a script that left that not equal to 1. I probably had wiggle importance 7 and shake accurary 4. Still not sure if that matters ;0 I was surprised that a shake/wiggle loop wouldn't converge. But now I've seen it for sure, I think I have seen it once or twice before. |
|
The options pop-up says that "# of iterations" affects both shakes and wiggles, but when I follow the code through it looks like RepeatContract call Contract and it ends up in fsl.MakeStable, which passes "iterations" to fsl.WiggleBackbone but not to fsl.Shake (which always uses 1). So "affects both shakes and wiggles" isn't exactly correct.