What I learned from Class Viewer
in [Java Programming]
|
From: rossum on 20 Jul 2008 15:46 On Sun, 20 Jul 2008 10:55:43 -0700 (PDT), JSH <jstevh(a)gmail.com> wrote: >On Jul 20, 4:12�am, rossum <rossu...(a)coldmail.com> wrote: >> On Sat, 19 Jul 2008 18:04:52 -0700 (PDT), JSH <jst...(a)gmail.com> >> wrote: >> >> >Like look at my Traveling Salesman Problem solution. >> >> How can it be a solution if there are instances of the TSP which it >> cannot solve? �At best you have a partial solution. >> > >You can always get distances between every node to the other by just >putting the nodes on a two-dimensional plane and saying one of the >weights is a distance measure. Not so. If the nodes can be plotted on a 2D plane then you are solving the Euclidian TSP. Not all instances of the general TSP are Euclidian TSP. For example distances on a sphere may not be plottable on a 2D plane. The instance may not even be metric. For example, the TSP with weights: |A B C --------- A|- 1 100 B| - 5 C| - is not metric as it does not obey the triangle equality and hence cannot be plotted on a 2D plane. In this case there is a toll bridge between A and C which means that you cannot use simple distance as an analogue for weight. > >So there are no instances for which my algorithm will not give an >answer. There are instances where the problem cannot be reduced to simple distance. rossum > >> >> >If it is correct, how many researchers in this area are suddenly >> >displaced by such a trivially easy algorithm? >> >> "If it is correct ..." �Do not get ahead of yourself James. �First you >> need to have a correct solution. �Start by amending your initial >> attempt so it can deal with Patricia's example that your current >> algorithm cannot solve. >> >> rossum > >She added in additional weights which she did not give values for, >though in her case those "weights" had to do with smuggling and gangs. > >A weight is ANYTHING that impacts the decision problem of whether or >not a particular path should be taken. > > >James Harris
From: Patricia Shanahan on 20 Jul 2008 15:59 rossum wrote: > On Sun, 20 Jul 2008 08:08:36 -0700, Patricia Shanahan <pats(a)acm.org> > wrote: .... >> The first step is to precisely define the problem it is designed to >> solve. Only JSH can do that. >> >> Until that is done it is impossible to know, for example, whether or not >> he is solving a problem whose decision form is NP-complete. It is also >> impossible to evaluate correctness without knowing what problem the >> algorithm is intended to solve. > From his insistence on "distance" rather than "weight" I would suspect > that he is trying to solve either the Euclidian TSP or the Metric TSP. My first guess would be 2-Dimensional Euclidean TSP. However, JSH does not seem to attach much significance to conventional definitions, so even if he agreed to that characterization I would not be sure that what he meant by it was the same as the definition in some reference book. Patricia
From: JSH on 20 Jul 2008 16:04 On Jul 20, 12:46 pm, rossum <rossu...(a)coldmail.com> wrote: > On Sun, 20 Jul 2008 10:55:43 -0700 (PDT), JSH <jst...(a)gmail.com> > wrote: > > >On Jul 20, 4:12 am, rossum <rossu...(a)coldmail.com> wrote: > >> On Sat, 19 Jul 2008 18:04:52 -0700 (PDT), JSH <jst...(a)gmail.com> > >> wrote: > > >> >Like look at my Traveling Salesman Problem solution. > > >> How can it be a solution if there are instances of the TSP which it > >> cannot solve? At best you have a partial solution. > > >You can always get distances between every node to the other by just > >putting the nodes on a two-dimensional plane and saying one of the > >weights is a distance measure. > > Not so. If the nodes can be plotted on a 2D plane then you are > solving the Euclidian TSP. Not all instances of the general TSP are > Euclidian TSP. For example distances on a sphere may not be plottable > on a 2D plane. The instance may not even be metric. For example, the My algorithm calls for straight line distance between nodes. Even on a sphere, there is a straight line distance even if it goes through the sphere as the algorithm doesn't care. It just wants the straight line distance between the nodes without regard to the surface or whether you can actually go that path. If the straight line distance goes through the middle of the earth, that's what the algorithm wants. There are details about the algorithm which are there for a reason. Deliberately ignoring things like a line between two points is just a waste of time. > TSP with weights: > > |A B C > --------- > A|- 1 100 > B| - 5 > C| - > > is not metric as it does not obey the triangle equality and hence > cannot be plotted on a 2D plane. In this case there is a toll bridge > between A and C which means that you cannot use simple distance as an > analogue for weight. A toll is a weight. Simplest thing is to just use the amount of the toll itself, though you may shift the weight depending on its impact--as people do in the real world. Issue here is that the academically stated problem throws away distance information, which, well, is stupid. People taught one way have an investment in that teaching: years of effort, sense of prestige with their school and their accomplishment, and social status to the academic world and its pronouncements. But they did the Traveling Salesman Problem wrong. Denial does not change the reality that distance information is key info that should not just be thrown away--even if a professor you admire greatly or a book you admire greatly, or some other person or persons you just trust taught you wrong. Being told that something valued is not worth as much as previously thought is a huge weight creating a giant cost for a particular path. By the TSP people will tend to go a different way. Your life is governed by the proper definition of the problem. Your life has always been governed by the proper definition of the problem, as you cannot escape it. The straight line distance from acceptance of the failure of what you were taught to the correct answer requiring distance information can be abstracted out in a space where the length of the path along acceptance for your case is, unfortunately, probably infinite as you are someone I do not expect to be capable of accepting reason over your emotional investment in the wrong knowledge. James Harris
From: Joshua Cranmer on 20 Jul 2008 16:53 JSH wrote: > Even on a sphere, there is a straight line distance even if it goes > through the sphere as the algorithm doesn't care. It just wants the > straight line distance between the nodes without regard to the surface > or whether you can actually go that path. If the straight line > distance goes through the middle of the earth, that's what the > algorithm wants. /me is forcibly reminded of the Mars travel thread on sci.math and does not want to repeat that. > There are details about the algorithm which are there for a reason. You are confusing me here. You seem to be simultaneously stating that distance is important and that it is not important at the same time (by virtue of your moon example, the point of which still eludes me). So, which is it? > Issue here is that the academically stated problem throws away > distance information, which, well, is stupid. You seem to be (re?)discovering some sort of analogue of bidirectional A* for TSP. If you recall the details of A*, you will note that it is correct as long as the heuristic only underestimates the distance. There are many cases--some which you yourself have given--where a blind distance metric does not guarantee an underestimation, thereby rendering such a heuristic invalid. > But they did the Traveling Salesman Problem wrong. No, TSP is essentially an esoteric question. In its most notable role, an NP-complete problem, distance becomes a redundant metric, since you define it then as the weights themselves. Cases where distance becomes involved in the TSP are special cases, labelled Euclidean-TSP, I believe (WP is down for me ATM). > By the TSP people will tend to go a different way. SENTENCE PARSE ERROR. > Your life has always been governed by the proper definition of the > problem, as you cannot escape it. As has been pointed out in sci.math (which is not so abhorrent a forum, you know), you can't simply take something (like pi) and redefine it as something else. You have to show that the two definitions are equivalent, at which point they are interchangeable. For example, I can't just blithely say that it's easier to solve 3-SAT than SAT and solve 3-SAT as a general solution. I have to show (which is actually quite easy) that all SAT problems are actually 3-SAT problems. Similarly, you can't use an annotated version of TSP to solve all TSP problems, merely because annotation is easier, you have to show that that the two are the same. > The straight line distance from acceptance of the failure of what you > were taught to the correct answer requiring distance information can > be abstracted out in a space where the length of the path along > acceptance for your case is, unfortunately, probably infinite as you > are someone I do not expect to be capable of accepting reason over > your emotional investment in the wrong knowledge. You know what, I would say more than 50%, probably in the range of 80-90%, of those who read your posts would claim the same foible in you. I've had a prolonged argument (to say debate would make me feel better but is probably less accurate) with a troll before and I'm sure no one here wishes the same thing to repeat again. Anyone else sensing a repeat of "Java and software piracy" or "Great SWT program"? -- Beware of bugs in the above code; I have only proved it correct, not tried it. -- Donald E. Knuth
From: JSH on 20 Jul 2008 17:49
On Jul 19, 6:06 pm, Joshua Cranmer <Pidgeo...(a)verizon.invalid> wrote: > JSH wrote: > > One of the reasons I bothered to create the open source project Class > > Viewer for Java was that I'd had years of arguing with math people on > > newsgroups which kind of happened by accident, as kind of as a lark I > > thought I'd ponder some math problems and talked about it. Yuck, what > > a mistake. > > > Math people ripped on me endlessly and told me to shut-up. > > FWIW, given my reading of the newsposts, most of the initial anger was > derived from perceived inadequacies in responding to perceived flaws or > inadequate specification in the algorithm--much like the questions > Patricia and I posed to you recently. Subsequent anger originated from a > positive-feedback cycle wherein you became more detached from your > detractors, inducing them to be even angrier (which is actually probably > not the right word here), which made you more detached, etc. Irrelevant to the crucial problem: proof didn't matter. > > I got angry and refused and found I LIKED pondering math problems they > > had claimed as their exclusive property and screw them. Over the > > years I learned many of them are idiots and they lie a LOT. It is > > scary how much they lie about, and there's nothing you can do about it > > and most people will not believe it's true because they think these > > math people are geniuses. > > ... and I would consider this a gross mischaracterization of the > situation. I don't believe anyone claimed FLT or factoring as "exclusive > property." Certainly, however, I don't think a cabal arose creating > great myths, but conspiracy theories do seem to be the rage these days. > To each his own. Do a search on "SWJPAM". I DID get published in a peer reviewed mathematical journal. The sci.math newsgroup mounted an email assault against my paper. No conspiracy theory there. It happened. The editors yanked my paper AFTER publication. No conspiracy theory there. A few months later the mathematical journal folded. No conspiracy theory there. It happened. The university which had hosted it, Cameron University, part of the Oklahoma state university system, removed all mention of the journal from its webpages. No conspiracy theory there. Want a link? Go to: http://www.emis.de/journals/SWJPAM/ The published editions of the journal were saved from loss by a European agency. No conspiracy theory there. > FWIW, as well, your postings here are not helping your case. There is nothing I can do to help my case!!! Do you not understand? These people have locked everything down. There is nothing that I can do. The truth does not matter. > > Math people lie all the time. > > My interpretation of this is best characterized as an attempt to > rationalize criticisms that many would consider to be valid, or, in > other words, to blame others for one's foibles. > > Hmm, I should really start venting outside of writing responses, as I > seem to be coming across as rather pompous for my tastes. > You have your head in the sand, and why not? I'm sure you're comfortable in your own way. It'd take pain and loss and desperation for you to care. That's the sad thing: no matter what people say, actually caring is usually about fearing loss. Comfortable people do not stop uncomfortable wars. James Harris |