QMC - What's it about?

[jimwah]

Quantum Monte Carlo, is a project based at The University of Münster, Germany.

The ability to accurately predict molecular structure and reactivity is of importance to the whole of chemistry and life sciences. Quantum Chemistry is a vital field of theoretical research into these phenomena. A very promising new Quantum Chemistry method is Quantum Monte Carlo (QMC). Our project Quantum Monte Carlo At Home (QMC@HOME) is dedicated to the further development of QMC for general use in Quantum Chemistry, to predict the structure and reactivity of molecules important to chemistry and life sciences.

I have quoted a post from the head honcho, Stefan Grimme (akgrimme) on the QMC forum (original thread here), on the subject of how Quantum Chemistry can be applied to current & future scientific processes...

Chemistry;
- calculations on model system for catalysts provide inside into catalysis
and help to design better catalysts; very important for the chemical industry,
therefore very important for our everyday life.
(A model system is a system where one leaves out all parts of minor importance
and takes into account only the very important regions of a molecular system.)

Biology;
- calculations on model systems for active sites of proteins
provide insight into biological processes;
one should keep in mind, that the theoretical background of e.g.
nowadays folding simulations are models often based on so called force fields
which are empirically fitted to specific classes of molecules
and therefore will not give good results for completely different systems;
quantum chemistry can help to greatly improve models like these.

Medicine;
- modelling of pharmaceuticals and there reactivity (e.g. host/guest systems)
provides insight on how these compunds operate and can give at least
good starting points for further research.

Material science;
- the molecular dimension becomes more and more important for compound development (e.g. nanoscience),
therefore modelling of such systems will become more and more important in the future.

After all I think, if you want to further develop fundamental theoretical models for biology and medicine you have to go from physics to chemistry to biology and further to medicine and here one very important point to start with is the application of physics to chemistry (this is Quantum Chemistry!), as in many parts (of course not all!) biology is based on chemistry and medicine on biology.

Optimised BIONC clients are no longer deemed acceptable & may be penalised; the introduction of 'fixed credit' WU's may negate any positive effect the optimisations had soon anyway.

Also, there was an issue with work units worth more than 1000 points not being credited. As far as I am aware this problem has been addressed, the WU credited, and I personally haven't seen a WU worth more than ~500 points in over a month.

Update - 21/09/06: Some larger work units are currently being distributed, and I can report that 3 WU over 1000 points have been credited without issue on my account in the last week.

Update - 08/01/07: Information regarding Optimised clients changed.

Useful Links:

QMC Homepage : http://qah.uni-muenster.de/
XS Team Page : http://qah.uni-muenster.de/team_display.php?teamid=467
Free-DC QMC stats : http://stats.free-dc.org/new/projpage.php?proj=qmc
More info & links here : http://qah.uni-muenster.de/scientific.php
Current Work Units : http://qah.uni-muenster.de/projectinfos2.php

[jimwah]

Project Admin @ QMC has made a statement in a news thread here regarding their, more aggressive, stance toward the use of 'optimized' clients:

A WORD ON CREDITS:
We, the developers of QMC@HOME, don't think it
to be 'fair play' to use 'optimized' clients,
that claim ten times the credit for the same work
(with no effect on calculation times)
and reserve the right to cut down credit
for clients that have claimed credit beyond
every reasonable limit.
This problem is hopefully solved with
'fixed credits' then.

To me, this is a clear message. I will be moving to a vanilla client. The introduction of fixed credit WU's should negate any effect the opti's have; but the guy's running the project sound very busy, so I'm pre-empting the move.

[[XC] riptide]

I must remember this then. Does this mean the benchmark goes out the window? One thing I hate so much about BOINC is the damn wacky benchmark.

[[XC]melymel]

akgrimme...... his name rings bells with the rosetta mess..... so long as they don't back date the credits fair enough it's stock from now on, but I am not getting my hopes up..... I hate boinc!!!!

[jimwah]

Nothing I've read indicates there will be any fiddling with stats, or zero'ing of credits obtained by users & teams so far. To do either, would be a minefield that could cause problems that the QMC team just don't have time for imo.

From the little I understand the BOINC benchmarks should make no difference in a fixed WU situation. Exactly how the value will be assigned to each unit, I don't know; perhaps ran on a benchmark machine first, in a F@H style

Whatever happens, I like the science behind QMC & I hope the new system is the solution that some people have been calling for; until then I run stock. Only got a few boxen on for now just to test the waters.

[[XC]melymel]

From what that flamebaiting twat of a QMC mod from boinc synergy was saying they were gonig for the brilliant approach of setting a large quorum so all those in the quorum get the average score of the 2 middle pc's with the anomolous results removed, same as what they do on WCG. As for the backdating credit issue I brought up; you're correct that they don't have time for the potential trouble so unfortunately they will likely just listen to the whiners thanks to that moderator, I think he already chased off PCZ from FDC with his points whinging so I don't hold much hope.... Why can't DC programs just make their own clients, it'd save so much trouble....

[jimwah]

Cutting down to the bone of the matter mely I see.

Why can't DC programs just make their own clients, it'd save so much trouble....

Yeah I'd agree that would eliminate a fair proportion of the problems, regarding 'cheating' anyway. As long as we have some teams to hand out a spanking to then I'm not overly concerned by what means my points are gathered, if it means running stock/quotum/whatever & laying down more power, so be it

I'm keeping an eye on my results, the first few were opti, the ones coming in now are not - first unit looks a bit under half the opti'd credit, maybe 40% Not so bad.

[[XC]melymel]

I'm keeping an eye on my results, the first few were opti, the ones coming in now are not - first unit looks a bit under half the opti'd credit, maybe 40% Not so bad.

I suppose that's a good thing as it should help slow down P3D some more!

[[XC] riptide]

Right. Lets someone report back stock or 5.5X. Need to keep a mental note of these things. So its a complete no no with opti's? Dead WU's or bad points?

[jimwah]

All I've wanted to see is a firm stance on whether opti's are acceptable at QMC, after much 'hinting' that they aren't, the statement above finally pins it down, therefore I've edited this sticky to fall in line, and have adopted the vanilla client myself. To be honest the opti's I've used certainly wouldn't group me with people claiming 10x credit, it's looking more like 2.0 - 2.5x so far (still waiting for more results here).

Whether or not any people will have credits taken away, and how that might be enforced, and whether other people will follow suit & go vanilla, remains to be seen