MathGroup Archive 2006

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Possible simple bug in NMaximize

On 6 Nov 2006, at 19:05, Andrzej Kozlowski wrote:

> On 6 Nov 2006, at 16:52, Nick Grishin wrote:
>> (* does not work *)
>> func = Hold[NIntegrate[t^3, {t, x, 0}]];
>> NMaximize[func, x]
>> (* works *)
>> func = Hold[NIntegrate[t^3, {t, 0, x}]];
>> NMinimize[func, x]
>> $Version
>> 5.2 for Microsoft Windows (June 20, 2005)
> A bug in NMaximize? Why? It behaves exactly as it should do with  
> this kind of input.  The problem is with MMinimize, because you  
> when you give it nonsensical input it actually produces output,  
> which unfortunately (by sheer accident),  looks as if it might be  
> right. But it is nonssense nonetheless (nonsense in - > nonsense  
> out) nonsensical, and  to convince you of it I will slightly change  
> your input:
> func = Hold[NIntegrate[t^3, {t, 10, x}]];
> NMinimize[func, x]
> {-2500.0000000000005, {x -> -0.0002833987269403981}}
> Now, does this look like it "works"?
> Andrzej Kozlowski

I have to admit that my example was pointless; it does also "work"  
just as much as your example works, although it should not. I agree  
that seeing this give what seem to be "right answers" might convince  
one that using Hold in this way is the right thing to do, but it is  
not. If there is a bug than it is in NMinimize and not in NMaximize  
and Hold should never be used in this sort of situation.

Andrzej Kozlowski

  • Prev by Date: Re: Evaluating integral with varying upper limit?
  • Next by Date: Re: Possible simple bug in NMaximize
  • Previous by thread: Re: Possible simple bug in NMaximize
  • Next by thread: Re: Possible simple bug in NMaximize