What I learned from Class Viewer
in [Java Programming]
|
From: JSH on 20 Jul 2008 18:28 On Jul 20, 12:59 pm, Patricia Shanahan <p...(a)acm.org> wrote: > rossum wrote: > > On Sun, 20 Jul 2008 08:08:36 -0700, Patricia Shanahan <p...(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 I don't like when people say something is right because of a definition, when it's obviously wrong. That strategy is to deny the truth using social constructs. Contradictions can be introduced with any real world implementation of a TSP algorithm that completely ignores distance information showing a flawed problem formulation. The cheat that people do, of course, is that they put the information back in when doing real world problems! But why take it out in the first place? Because some academics made a definition? Sorry, but that is just willful stupidity which would not occur if you had to do it the right way or fail. The reason you can argue this point is that your life doesn't depend on the right answer, and nothing important enough to you depends on the right answer, so you have the luxury of stupidity in this regard. Stupid in the real world is often about the luxury to be willfully wrong. Distance information is part of the TSP whether you acknowledge it or not. James Harris
From: Patricia Shanahan on 20 Jul 2008 19:42 JSH wrote: > On Jul 20, 12:59 pm, Patricia Shanahan <p...(a)acm.org> wrote: >> rossum wrote: >>> On Sun, 20 Jul 2008 08:08:36 -0700, Patricia Shanahan <p...(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 .... > Distance information is part of the TSP whether you acknowledge it or > not. .... No, in my unfortunate ignorance I don't know in exactly what way distance information is part of the TSP. It certainly is not part of the version of TSP in the reference books I do know about. However, I am always willing to learn. Why not just state what you consider to be the correct definition of TSP, the one you are using in your algorithm? That way, we would *know* what you meant by TSP, rather than having to guess. Patricia
From: John W Kennedy on 20 Jul 2008 19:51 JSH wrote: > Contradictions can be introduced with any real world implementation of > a TSP algorithm that completely ignores distance information showing a > flawed problem formulation. Do you /still/ not understand? It's not a question of "ignoring" the distance. In the real world, it's a question of evaluating /everything/. Distance, is a factor, but it's not the only factor, and it's not a direct factor at all, only indirect. Distance affects the time (but not necessarily one-for-one). It affects fuel costs (but not necessarily one-for-one). But there are other considerations unrelated to pure distance. What about tolls? What about bottlenecks, especially bridges and tunnels near big cities? What about the fact that roads don't go in straight lines, and also have uphill and downhill sections (and, no, in practical calculations, those don't cancel out)? What about wear and tear and depreciation? What about air flights and trains being available between some cities and not others? What about waiting time in airports and train stations? What about hotel rates? And, because some of these things are measured in hours, and some are measured in dollars, you have to decide how much the traveler is worth in dollars per hour, so that it all reduces to one number (either dollars or hours will do for the final result). In your nothing-but-distance version of the problem, all the values work out in neat numbers where all the diagonals follow the Pythagorean theorem. But in the real-world situation, that isn't true. The shortest distance is not normally the fastest, and the fastest is not normally the cheapest, and the cheapest is not normally the shortest. If you want an example, just get any kind of GPS or other route-planning software that has a fastest/shortest option and compare the results on any non-trivial journey. The TSP problem, in general, has to work where the numbers are of these messy kinds. As Patricia says, even if you /do/ have an algorithm, it doesn't solve the standard TSP, because it only solves the 2-D Euclidean simplification, where it is safe to assume that ABC >= AC. -- John W. Kennedy A proud member of the reality-based community.
From: Lew on 20 Jul 2008 20:16 JSH wrote: > Sorry, but that is just willful stupidity which would not occur if you > had to do it the right way or fail. The reason you can argue this > point is that your life doesn't depend on the right answer, and > nothing important enough to you depends on the right answer, so you > have the luxury of stupidity in this regard. When you resort to name-calling you prove two things. First, that your idea does not stand on its merits. Second, that you are a troll, JSH. Plonk. -- Lew
From: Joshua Cranmer on 20 Jul 2008 22:23
JSH wrote: > There is nothing I can do to help my case!!! And you've ignored the one saving grace that will make people actually listen to you: /respond/ to criticism. Merely sitting on a pedestal and shrilly shouting "YOU'RE ALL WRONG! YOU'RE ALL STUPID IDIOTS DWELLING IN YOUR OWN QUAGMIRE OF EXCREMENT INCAPABLE OF SEEING THE TRUTH! I'M THE ONLY SANE PERSON HERE!" is not going to win you any sympathy or friends. If you go back over your first thread about TSP again and read carefully, you'll notice that people didn't mind your algorithm until you started screaming in objection to Patricia's classification of your algorithm as not solving the TSP. One thing to keep in mind is that Patricia and Lew (among many others in this newsgroup, among whom I would not count myself) have established themselves as experts in this newsgroup, so their criticism is not the idle thoughts of a neophyte, but it is instead the result of expertise. This expertise has not been established by yourself, so deferring to masters in a field is a wise recommendation. In short, if you want to redeem yourself, all you have to do is the following: 1. Stop screaming and complaining that the world is against you. 2. Show a step-by-step example of how the algorithm works for a 5- or 6- node problem (at least) in a clear manner, referring to diagrams for extra clarity. Use only the information provided in the standard NP-complete (or NP-hard, if you choose) version of the problem. 3. Explain in a detailed manner why you believe the algorithm works for all cases. 4. Evaluate your algorithm to determine which cases would most likely break the algorithm. If you complete steps 1-3, you'll probably find willing volunteers to help with this step. 5. Be prepared to assuage our doubts about your algorithm. Most notably, if we believe there is a breaking step, be ready to defend why the step does not break. If we provide an example that breaks your algorithm, evaluate why your algorithm fails and revise it with that information in mind. The most important step here is number 5. Trust me, being polite in feedback and in responses to feedback opens a lot of otherwise closed doors. > You have your head in the sand, and why not? You obviously know little about me. I read two daily newspapers--The Washington Post and The Christian Science Monitor--as well as the weekly Economist. I keep up very well on current affairs, and am generally observant in my surroundings. But I'll forgive you for not knowing the depth to which I look outside the shell of the area in which I work and live. > It'd take pain and loss and desperation for you to care. And here I do not forgive you for not knowing my qualities (well, I do, but that's another story). If you look through my record of postings in sci.math and c.l.j.p, you'll find that I have far from tried to demonize you but instead tried to help you along. And outside of the record of newsgroups, you'll find that it is in my nature to always give someone the benefit of the doubt and try to work along to a solution to a degree beyond which many have given up. It is my nature to always care. I think you'll find that assuming the worst in a person tends to bring out the worst, but assuming the best tends to bring out the best. If I really didn't care, after all, I wouldn't be posting, now would I? > Comfortable people do not stop uncomfortable wars. You're in the wrong newsgroup. Look over in talk.politics, my friend. -- Beware of bugs in the above code; I have only proved it correct, not tried it. -- Donald E. Knuth |