HP Forum Archive 14

[ Return to Index | Top of Index ]

HP solver has some bugs Message #1 Posted by will on 28 Feb 2005, 12:41 p.m. try to find a root for 25*x-5*ln(x)=0 using the hp49g solver...

Re: HP solver has some bugs Message #2 Posted by Mike (Stgt) on 28 Feb 2005, 2:12 p.m., in response to message #1 by will Plot it from x=0 to .5 and tell us what you expect. Ciao.....Mike

Re: HP solver has some bugs Message #3 Posted by Gene on 28 Feb 2005, 2:40 p.m., in response to message #2 by Mike (Stgt) Well, it shouldn't return 0.2 something. :-) Gene

Your example does not suggest a problem... Message #4 Posted by Karl Schneider on 1 Mar 2005, 12:41 a.m., in response to message #1 by will "will" -- My 33S also can't find a real root to your equation. I suspect that neither can my 15C, 28C, 32SII, 34C, 41C*/Advantage, 42S, 48G or 49G. Mike and Gene might be coy, but I think you really meant, 25*x+5*ln(x)=0, for which x = 0.26534 -- KS Edited: 1 Mar 2005, 12:42 a.m.

Re: Your example does not suggest a problem... Message #5 Posted by Axel on 1 Mar 2005, 9:26 a.m., in response to message #4 by Karl Schneider y=25*x-5*ln(x) has no real solution, whereas y=25*x+5*ln(x) can be solved by HP Solver (tried on 32sii and 40g, solution is 0.2653449...) Regards, Axel

World's all solver has a bug - or not??? Message #6 Posted by Tizedes Csaba [Hungary] on 1 Mar 2005, 3:02 a.m., in response to message #1 by will My favourite, I posted it below, but NOBODY can tell me, what is the problem in the symbolic solvers!!! Heyhooo Mathematica, Maple, etc programmers!!!! Here is the little problem, what can solve an 10 yrs student, but your 10 yrs old symbolic progs CANT!!!!! Here it is: 1 1 3*x + ----- = 15 + ----- (x-5) (x-5) And the solution is that: "No solution!" All of softvares gives x=5 that is not root!!! Feed your head! Csaba

you are right and you are not Message #7 Posted by Mike (Stgt) on 1 Mar 2005, 4:23 a.m., in response to message #6 by Tizedes Csaba [Hungary] True, your function at x=5 is _not_ defined. But (!) the root is a "hebbare Lücke" (sorry, I know only the German word; about "repealable or liftable gap") as the values of the function and its derivative at x=5-0 and x=5+0 are the same. But I am quite sure that no CAS cares about adjacence of gaps in a function. Ciao.....Mike

Re: World's all solver has a bug - or not??? Message #8 Posted by will on 1 Mar 2005, 10:51 a.m., in response to message #6 by Tizedes Csaba [Hungary] Interesting posts...the bug is that the HP solver shows a false answer around 0.2 something which is not a root...in fact there is not real answer... about the problem that csba told us is quite important avoid divide by 0.

Re: HP solver has some bugs Message #9 Posted by Cyrille de Brébisson on 1 Mar 2005, 1:50 p.m., in response to message #1 by will Hello, Actually, the solver works perfectly well. the problem is the user interface which returns 0.2 (and some change) but without displaying the appropriate error/information message: "Extremum" If you try the "OLD" version of the equation solver UI: hold red shift, 7, ROOT '25*X-5*LN(X)' ENTER green shift EQ, SOLVR green shift X, you will get 0.2 and change and the message Extremum saying that the solver found an extremum, but not a solution Cyrille

Actually... Message #10 Posted by John on 1 Mar 2005, 7:21 p.m., in response to message #9 by Cyrille de Brébisson All one has to do on the new solver is press the menu label "INFO" and it will indicate that the 0.2 was found as an Extremum. This is ALWAYS a good idea and is proper use of the solver. One should never! accept a numerical result from a solver without knowing this type of additional information. Always press INFO to find out how the result was obtained! Gene

which version/revision? Message #11 Posted by Mike (Stgt) on 2 Mar 2005, 9:17 a.m., in response to message #9 by Cyrille de Brébisson Aha! HP listens to the grapevine -- Bon jour CdB! :) Could you pls tell us on which ROM version the "old" UI is available? I tried (i) Version HP49-C, Revision #1.18 (no difference between red-shift 7 and hold-red-shift 7), (ii) Version HP49-B, Revision #1.19-6 (same as i), and (iii) Version HP49-C, Revision #1.24 but there I was not able to reproduce your example. Suppose you are on a HP-49G+plus. BTW, sorry if I ask for info that is described elswhere and I did not find it: what is the meaning of version letters on the HP49G? It seems not to correspond with the revision numbers. B=Beta and C=produCtion? TIA.....Mike Edited: 2 Mar 2005, 9:18 a.m.

Re: HP solver has some bugs Message #12 Posted by Bernardo Vargas on 2 Mar 2005, 5:59 a.m., in response to message #1 by will The problem is that SOLVE starts with the value x=0, and then returns immediately an error message because of the impossibility of ln(0). In this case the SOLVE program on the MATH-1 module for the HP41 works better, since it asks first for two GUESSES for the solution. This way you can avoid the ln(0) by entering 0.1 or something like that. Then the program runs for some seconds and gives the answer ROOT IS 0.265344933. The manual of the HP32SII has an interesting appendix telling how SOLVE works, and explains some of its limitations. And let me remind you that EVERY numerical algorithm has its limitations. So is mathematics, sorry.

[ Return to Index | Top of Index ]

Go back to the main exhibit hall