Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.soft-sys.math.mathematica > #15451 > unrolled thread

very odd failure of Solve

Started byAlan <alan.isaac@gmail.com>
First post2013-09-15 11:03 +0000
Last post2013-10-16 08:46 +0000
Articles 7 — 5 participants

Back to article view | Back to comp.soft-sys.math.mathematica


Contents

  very odd failure of Solve Alan <alan.isaac@gmail.com> - 2013-09-15 11:03 +0000
    Re: very odd failure of Solve Richard Fateman <fateman@cs.berkeley.edu> - 2013-09-16 08:03 +0000
      Re: very odd failure of Solve "Dr. Wolfgang Hintze" <weh@snafu.de> - 2013-09-18 01:30 +0000
        Re: very odd failure of Solve Richard Fateman <fateman@cs.berkeley.edu> - 2013-09-18 10:11 +0000
      Re: very odd failure of Solve danl@wolfram.com - 2013-09-18 10:12 +0000
        Re: very odd failure of Solve Alan <alan.isaac@gmail.com> - 2013-09-22 06:21 +0000
        Re: very odd failure of Solve Itai Seggev <itais@wolfram.com> - 2013-10-16 08:46 +0000

#15451 — very odd failure of Solve

FromAlan <alan.isaac@gmail.com>
Date2013-09-15 11:03 +0000
Subjectvery odd failure of Solve
Message-ID<l1446c$psj$1@smc.vnet.net>
Setting an irrelevant parameter to 0 baffles Solve. Why?
Thanks,
Alan Isaac

$Assumptions =.
ClearAll[f1]
f1[x_] := s*x^\[Alpha] - (a + b + c)*x
Solve[f1[x] == 0, x]  (* Solve works *)
Solve[(f1[x] /. {b -> 0}) == 0, x]  (* Solve fails *)

(* adding these assumptions does not change things *)
f1assume =  (1 > s > 0 && k >= 0 && 1 > \[Alpha] > 0 && a >= 0 && b >= 0 && 
   1 > c > 0);
Assuming[f1assume, Solve[f1[x] == 0, x]]  (* Solve works *)
Assuming[f1assume,  Solve[(f1[x] /. {b -> 0}) == 0, x]] (* Solve fails *)

[toc] | [next] | [standalone]


#15454

FromRichard Fateman <fateman@cs.berkeley.edu>
Date2013-09-16 08:03 +0000
Message-ID<l16e0j$2g3$1@smc.vnet.net>
In reply to#15451
On 9/15/2013 4:03 AM, Alan wrote:
> Setting an irrelevant parameter to 0 baffles Solve. Why?
> Thanks,
> Alan Isaac
>
> $Assumptions =.
> ClearAll[f1]
> f1[x_] := s*x^\[Alpha] - (a + b + c)*x
> Solve[f1[x] == 0, x]  (* Solve works *)
> Solve[(f1[x] /. {b -> 0}) == 0, x]  (* Solve fails *)
<snip>

Running Reduce[ {%==0}, {x} ] on either equation seems to go into
an infinite loop.  That maybe be an independent bug, though.

I expected that somehow fiddling with the variable names would
do something, and that the ordering of a,b,c, alpha was critical.
Out of curiosity I tried a few variants to generate a better
hypothesis, but ran out of, um, curiosity.


[toc] | [prev] | [next] | [standalone]


#15463

From"Dr. Wolfgang Hintze" <weh@snafu.de>
Date2013-09-18 01:30 +0000
Message-ID<l1avne$cvi$1@smc.vnet.net>
In reply to#15454
Am Montag, 16. September 2013 10:03:31 UTC+2 schrieb Richard Fateman:
> On 9/15/2013 4:03 AM, Alan wrote:
> 
> > Setting an irrelevant parameter to 0 baffles Solve. Why?
> 
> > Thanks,
> 
> > Alan Isaac
> 
> >
> 
> > $Assumptions =.
> 
> > ClearAll[f1]
> 
> > f1[x_] := s*x^\[Alpha] - (a + b + c)*x
> 
> > Solve[f1[x] == 0, x]  (* Solve works *)
> 
> > Solve[(f1[x] /. {b -> 0}) == 0, x]  (* Solve fails *)
> 
> <snip>
> 
> 
> 
> Running Reduce[ {%==0}, {x} ] on either equation seems to go into
> 
> an infinite loop.  That maybe be an independent bug, though.
> 
> 
> 
> I expected that somehow fiddling with the variable names would
> 
> do something, and that the ordering of a,b,c, alpha was critical.
> 
> Out of curiosity I tried a few variants to generate a better
> 
> hypothesis, but ran out of, um, curiosity.

Still a bit curious I tried

1) the number miracle

Solve[s*x^\[Alpha] - (a + b + c)*x, x] (* Solve works *)
Solve[s*x^\[Alpha] - (a + b )*x, x] (* Solve fails *)
Solve[s*x^\[Alpha] - (a )*x, x] (* Solve works *)

oops, size matters (1 and 3 is okay, 2 not)

2) the simplest case

The most simple case of this type in which Solve fails seems to be

Solve[x^E - 2*x == 0, x]

Solve::nsmet: This system cannot be solved with the methods available to Solve. >>

You can safely replace the irrational E by the algebraic Sqrt[2] or other constants apart from rationals. Solve will fail as well.

Competition is open to find the simplest case Solve can't do with.

Best regards,
Wolfgang


[toc] | [prev] | [next] | [standalone]


#15466

FromRichard Fateman <fateman@cs.berkeley.edu>
Date2013-09-18 10:11 +0000
Message-ID<l1bu93$fn8$1@smc.vnet.net>
In reply to#15463
On 9/17/2013 6:30 PM, Dr. Wolfgang Hintze wrote:

....

Reduce[{x^a-x==0},x]
FullSimplify[%]

[toc] | [prev] | [next] | [standalone]


#15467

Fromdanl@wolfram.com
Date2013-09-18 10:12 +0000
Message-ID<l1buai$fnj$1@smc.vnet.net>
In reply to#15454
On Monday, September 16, 2013 3:03:31 AM UTC-5, Richard Fateman wrote:
> On 9/15/2013 4:03 AM, Alan wrote:
>
> > Setting an irrelevant parameter to 0 baffles Solve. Why?
>
> > Thanks,
>
> > Alan Isaac
>
> >
>
> > $Assumptions =.
>
> > ClearAll[f1]
>
> > f1[x_] := s*x^\[Alpha] - (a + b + c)*x
>
> > Solve[f1[x] == 0, x]  (* Solve works *)
>
> > Solve[(f1[x] /. {b -> 0}) == 0, x]  (* Solve fails *)
>
> <snip>
>
>
>
> Running Reduce[ {%==0}, {x} ] on either equation seems to go into
>
> an infinite loop.  That maybe be an independent bug, though.
>
>
>
> I expected that somehow fiddling with the variable names would
>
> do something, and that the ordering of a,b,c, alpha was critical.
>
> Out of curiosity I tried a few variants to generate a better
>
> hypothesis, but ran out of, um, curiosity.

It does seem to involve ordering. If instead of "a+c" one has "y+c" then Solve handles it. The underlying issue is some simplifying that produces, or not, a "nice" form of intermediate expression. This was also recently reported in the Mathematica stackexchange forum.

http://mathematica.stackexchange.com/questions/25182/variable-naming-changes-everything

I hope to find time to investigate further, some day.

Daniel Lichtblau
Wolfram Research

[toc] | [prev] | [next] | [standalone]


#15505

FromAlan <alan.isaac@gmail.com>
Date2013-09-22 06:21 +0000
Message-ID<l1m29c$ca4$1@smc.vnet.net>
In reply to#15467
On Wednesday, September 18, 2013 6:12:34 AM UTC-4, da...@wolfram.com wrote:
> It does seem to involve ordering. If instead of "a+c" one has "y+c" then Solve handles it. The underlying issue is some simplifying that produces, or not, a "nice" form of intermediate expression. This was also recently reported in the Mathematica stackexchange forum.
>
> http://mathematica.stackexchange.com/questions/25182/variable-naming-changes-everything
>
> I hope to find time to investigate further, some day.


Thanks for the link. Looking forward to a resolution.
Alan Isaac

[toc] | [prev] | [next] | [standalone]


#16064

FromItai Seggev <itais@wolfram.com>
Date2013-10-16 08:46 +0000
Message-ID<l3ljpo$jt9$1@smc.vnet.net>
In reply to#15467
> On Wednesday, September 18, 2013 6:12:34 AM UTC-4, da...@wolfram.com wrote:
> > It does seem to involve ordering. If instead of "a+c" one has "y+c" then
> > Solve handles it. The underlying issue is some simplifying that produces,
> > or not, a "nice" form of intermediate expression. This was also recently
> > reported in the Mathematica stackexchange forum.
> >
> > http://mathematica.stackexchange.com/questions/25182/variable-naming-chan
> >ges-everything
> >
> > I hope to find time to investigate further, some day.
>
> Thanks for the link. Looking forward to a resolution.
> Alan Isaac

I'm please to report that this particular issue is fixed in our internal build, 
so the next release will resolve it.

-- 
Itai

Itai Seggev
Software Technology

[toc] | [prev] | [standalone]


Back to top | Article view | comp.soft-sys.math.mathematica


csiph-web