Author: Robert Hyatt
Date: 08:35:29 09/06/02
Go up one level in this thread
On September 06, 2002 at 09:57:09, Rolf Tueschen wrote: >On September 06, 2002 at 09:11:07, Vincent Diepeveen wrote: > >>Committing fraud, even falsifying (and not even remembering it >>for christ sake) the 2 most important tables >> - total node counts >> - time table > >Your terminology is out of order. But I won't play coach again. All I can say is >that your understanding of the basic fundaments in science is very weak. >Falsifying at least is not what you could ever prove for these tables. Let me >make a sarcastic comment. You are sinking deeper and deeper into the moor... If >you only could see that alone you'd never could come out of it. If you only >could pause and say, well, then just give me a straight explanation for what I >see, because I have no idea why something like that could happen. Then it could >be much easier. But now you are headlessly searching for always new variations >of the same accusation. And the irony is that the accusation is false! Did you >mumble the word? Ok, I'll try it a last time. > >> >>This in an official article, which is one of the things you have to >>try to do as a professor. > >JICCA is the journal of a private association. Official for CC but not official >in the sense that a member of a university had to fear something. In this case I >think the journal itself had more interest in the paper than Bob needed it for >his career. And keep in your memory that Bob's findings are by no means hurt by >your table question. The reason why it seems so difficult for you to understand >that is exactly the lack I described in your 'basics'. Vincent, you simply can't >judge here what is most important, important, less important and completely >uninteresting. To show you a difference. As you know I accused the DB2 team of >unscientific behaviour to Kasparov, who was their client so to speak. Such a >"pressure" put on your client will always hurt your results. So, normally you do >your best to keep a good relationship with all your clients. This is so basic, >that even in the contract with IBM, scientists should always secure this, if for >instance IBM had tried to urge them somehow. Then they should have said, that >this way the whole event would be destroyed. Hence after game two the whole >event was heavily biased. Of course from a limited (not very scientifical) >standpoint of the factual the result is a win for DB2. And that is how it's seen >among laymen. > >Here we have a completely different question. If you only could show that the >numbers Bob reported were completely out of range, then you had a good argument. >But here the stories behind the publication, the loss of the original data and >the callings from the journal itself let Bob make a little mistake. But not a >mistake which meant anything false or contrary to the original results. Can you >understand that? Ask further questions. > > >>So you show falsified results to the other >>scientists about your data. > >No. He didn't change the results but presented them. But the presentation was >not cheating the scientists or you! If you had understood what Bob had told you >by email and ICC conversation! How can you be so stubborn to realise the >possibility of misunderstanding on your side. > > >> >>If the only way of getting the tables is by inventing numbers yourself, >>then do not give the tables. > >But the found results were the core or the main cause of the interest of the >JOurnal! > > >> >>A result from the time table IS the speedup number, and not vice versa >>as you post here. You do not get a speedup number from heaven. >> >>You CALCULATE it based upon the search times. >> >>Automatic processes sometimes round off numbers, that is not the >>case here. So the conclusion is very obvious. > >I'm sure you'll get in detail answers on the technology questions, but I can >only hope that you can understand the different levels of importance. But a fast >understanding is required. You are very redundant actually. > >Rolf Tueschen Let me add a note about the term "referee". I include two groups in that category. I won't name names, but: 1. When I write something, I generally get a couple of outside folks to go over it first, with the idea that they are acting like referees to see if they see wording problems, inconsistent statements, data errors, and the like. IE for the JICCA bitmap paper, I sent copies to several for comments before it made it to the icca and their referee process. 2. After submission, there is always a referee process, in general. Some are good, some or not. Some want no changes, some want lots of changes. The most common referee complaint I get is about a missing reference or two, which does sometimes "identify" the referee of course, if you get my meaning. :) But in any case, it is _not_ an adverserial relationship, it is more about making the paper clearer, more accurate, and, of course, almost _always_ shorter. :) So the referees were not demanding that I add something. And it wasn't a case of saying "hell no". It is very much like working with an anonymous group that _really_ does want to make the article read as well as it can read. I do this regularly for conferences, journals (including the JICCA) and even chapters of books. I never consider it my task to tear something apart. Rather it is my task to try to help the author improve it. I've done this for books I use in class, for example. I don't want to leave the impression that either journal or personal referees demanded something be added. At worst they make "strong suggestions" that when looked at carefully, usually make some kind of sense. Yes, I have problems with Jaap as the editor. He likes "colour". I like "color". I send him an article with "color" everywhere, the proof comes back with "colour" everywhere. I change it and send it back. :) of course, he gets the last "cut" so I end up looking like a Brit when I am not. :) But we both laugh about that and go on. (there are other things like "organisation" vs "organization" and the like, that drive me bananas also...) So this "data" issue is not a demand by them, nor a "I'll show 'em by producing it in an easy way" by me... It turned into a matter of "how can I get there from here?" and for nodes, there was no choice. Probably for times there was no choice due to lost data logs, but that is not something I specifically recall although the data is convincing that it was the case. In hindsight, a simple one page printout of all the test times would have solved this, although I am sure Vincent would have jumped right on the node issue since they _had_ to be extrapolated. But then two of the three tables would have been more consistent. However, my files (paper files) are huge. I have hundreds of kopec test runs for Cray Blitz. Thousands for my dissertation. I just didn't think to print that simple one page summary out, and the rest is history. For the current "study" (since I will probably write up the current crafty SMP algorithm since it is not to difficult to understand and relatively easy to implement) I have already put the data files on a second machine. Hopefully at least one person will download them so that we have a 3-way backup, in case this happens again. :) This was the first time I have had a massive data loss like that, and I plan on it being the last, since I now have my primary files (crafty, etc) on at least two different machines, one with raid-5 disk. It was very painful, and losing the DTS article data was insignificant compared to losing everything else. Electronic copy of my dissertation paper itself, copies of other papers. All the old cray blitz source code versions. All the log files for every game we had played in an ACM or WCCC event since 1976. Old email. Old test positions. Most was totally irreplacable. But it did teach me a valuable lesson... you can't trust a backup... :) For the endgame tables, I have multiple copies. All the 3-4-5 piece files are on a set of CDs locked away in my office. Another set is out of the US completely. I plan on backing up all the 6 piece files onto DVD once I get this damned EIDE DVD writer to work under linux... which requires SCSI emulation and a lot of other nonsense. It was unfortunate this caused such an uproar, mainly by one person. But it did, and it can't be undone. I don't see any "fraud" whatsoever, and while I wish (with hindsight) that we could have used real data, we couldn't, we didn't, and I can live with that... And I'll simply take more stringent safeguard actions in the future. You notice that all the crafty-goes-deep data is still available. As I have multiple copies saved away so that if this comes up there, we will have real data to fall back on... so I _do_ learn... :) I had already "learned" before this fiasco erupted. :)
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.