|How much longer will it be? |
I've hoping this one will finally be able to beat PhaseRotC's current best of 4.091798%, as I've put more power in the accelerator (actually, made it longer) and added some components at double-frequency to do more detailed tweaking of the time/energy phase space distribution.
|I'd volunteer to push in with some manual tweaking, but the parameters are too much to be read in with CSV format by Excel. If the ";" could be replaced by a CRLF, I could read them in vertically. I've done this earlier with a hex-editor and it was a PITA, so if someone could write a small proggy for this conversion it would be easier to handle this bunch of data.|
|Herb, I just used Microsoft Word to do this. Most word processors will let you replace a character (semicolon) with a Line Break. With Word, you have to pick "Line Break" out of a drop-down box at the bottom of the window. (it is just a Control-L )|
I'm not sure which file you want to work on, but the 100 sample file became 161 pages long after replacing the semicolons.
|Didn't I make a utility called Results2CSV (see muon front page) that does this sort of thing for you? Or are the results still too wide from that?|
|just use a text editor like notepad like a MAN herb|
|Wow, I hadn't thought about Word, finally a useful function in it I knew it worked in Wordperfect but that one is gone long time |
The results for PhaseRotD make up some 500 items, while Excel can read in only 256 columns. The results2csv tool is fine but it doesn't help with Excel at this width.
New statistical adventures ahead
|Maybe I should modify it to either split the table or output the transpose of the table, maybe as a commandline option?|
I have on one occasion had to use Word for a technical function, simply because it was the only program I had that could replace special characters like carriage returns by other things!
|OK an improved version has been uploaded here. Run it with no commandline arguments to see the usage.|
|The transpose as parameter would be most appreciated if possible.|
I did it with a hex editor before, but it was a 3 step operation, where the replace function didn't store the input, no joy duh.
|PhaseRotD is being a bit slow to go positive, though it is climbing steadily. I'm thinking of dropping in a semi-manual-designed seed. Should I wait?|
|Although I threw in a number of manual tweaks, it didnt't behave as all other lattices before. DecayRotB behaves much simpler. With PhaseRotD I rarely catch an improvement by manual adjustments, while with DecayRotB nearly every second trial shows an improvement or at least a trend where next to go to. I suspect as long as we are in the negative, there are no real trends yet obvious to detect, however I would give it another week to become positive.|
While I'm at it, we could have maybe a good improvement if we were able to direct the trial types a bit more. I usually churn manually on one parameter several times with interpolation and extrapolation when it is showing a trend and then finally optimize it closer and closer. If a new top design shows up I recheck it for the previously optimized parameters, this usually leads directly to the next improvement step in the stairway.
I won't have too much time for this over this next weeks, but I thaught already about a logic how to automaticaly generate TEST designs following some principal consequential rules. Such a logic would provide also TEST queues which could be distributed to clients on demand base, this would speed up the design improvements dramatically.
I'll come back on this matter on Wednesday, have a short mission to do now.
|I think it's about to do it soon - that -0.04 result could have been a re-checked mix of positive numbers with things in the range ~-0.1.|
|I'm observing the first positive runs in my queue, although the sum still goes negative. The best -0.012727 clearly showed better trends of some parameters now. |
I'm setting up now a huge queue for running unattended over the weekend now, hopefully some will be positive when I'm back on Monday.
|My client just got positive today (0.011823), but it takes quite long to give a breakthrough at the moment. From -0.10 to -0.0x went in huge steps but it seems my PC is slacking right now .|
|With the return of the stats, it shows that [DPC] Xanathorn was the first positive with 0.016010.|
|Congrats, going to check what finally made the turn|
|Thanks . I got like 200+ results positive so far but it seems my client is kinda stuck right now, but looking at the best so far list my client ain't the only one . Let's hope it's gonna catch some speed soon..|
|"Stuck"? Try the new 4.43d version of Muon1. Something in the optimiser was just a bit broken, so 4.43 has been behaving sub-par.|
|"Stuck" as in not getting better results, not stuck between the door or something like that...|