0
$\begingroup$

This is unconstrained optimization. First Order Necessary Condition (FONC) gives me two equations that are contradicting each other. x1 + x2 = 5 and x1 + x2 = -5. What does this imply?

$\endgroup$
2
  • $\begingroup$ Too little information. Describe the whole problem in more detail. $\endgroup$ May 3, 2014 at 22:50
  • $\begingroup$ Find maximum or minimum of a given function. I have \nabla f. Hessian is [-2, -2; -2, -2] $\endgroup$
    – user100503
    May 4, 2014 at 0:17

1 Answer 1

1
$\begingroup$

It's very difficult to answer this question without more information. However, the function

$f(x)=-x_{1}^{2}-x_{2}^{2}-2x_{1}x_{2}+10x_{1}-10x_{2}$

has the Hessian you specified, and if you set $\nabla f(x)=0$, you get the two equations that you've provided in your posting. This particular function has no local min/max points.

It's also possible that you've made an error in computing $\nabla f(x)$ or the Hessian.

$\endgroup$
3
  • $\begingroup$ This is exactly the function that I have. So looks like my \nabla and Hessian are correct. How do you find if this has a minimizer of maximizer? I tried the usual First Order Necessary Condition but that does not give me any points. How do you conclude that there are no maximizers or minimizers? Thank you very much. $\endgroup$
    – user100503
    May 4, 2014 at 3:30
  • $\begingroup$ If you fix $x_{1}$, then consider $f$ as a function of $x_{2}$ only, you'll see that the maximum of $f$ keeping $x_{1}$ constant occurs at $x_{2}=-5-x_{1}$ and that the maximum value is $20x_{1}+25$. Thus we can increase $x_{1}$ towards $+\infty$, move $x_{2}$ to $-5-x_{1}$ and keep getting larger and larger values of $f$ without bound. This is not unusual behavior- it's quite common for indefinite quadratic forms to have no maximum or minimum. $\endgroup$ May 4, 2014 at 3:39
  • $\begingroup$ At a different level, you should understand that because there are no solutions to the FONC and $f$ is smooth, then there are no minimizers or maximizers. $\endgroup$ May 4, 2014 at 3:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.