My book states:

"It is also true that if $x^*$ is an interior point and:

  • a global maximum of $f$ , then $d^2f(x^*)$ is negative semi-definite.
  • a global minimum of $f$ , then $d^2f(x^*)$ is positive semi-definite.

But, it is not true that if $x^*$ is a critical point and $d^2f(x^*)$ is negative (positive) semidefinite, then $x^*$ is a local maximum (minimum).

However consider the function: $x^4+x^2-6xy+3y^2$ .

It has a global minimum on $(x,y)=(1,1)$ and $(x,y)=(-1,-1)$ .

However the Hessian at this points, does not seem to be semi-definite.

Can anyone help me?

  • $\begingroup$ what do you think is the matrix of second partial derivatives of your function? $\endgroup$ – Will Jagy Aug 7 '14 at 20:59
  • $\begingroup$ I see your deleted second "answer." Yes, a definite hessian is also semi-definite. $\endgroup$ – Will Jagy Aug 7 '14 at 22:47

I would say it is the Hessian ...

Which shall be equal to:

\begin{matrix} 12x^2 +2 & -6 \\ -6 & 6 & . \end{matrix}

Which is positive definite for the solution at hand, right?

  • $\begingroup$ yes, positive definite at $(1,1)$ and $(-1,-1)$ which are local (and global) minima, and indefinite at $(0,0)$ which is a saddle point. Maybe your trouble is the prefix "semi." The word semidefinite includes definite. $\endgroup$ – Will Jagy Aug 7 '14 at 21:21
  • $\begingroup$ Willt, thanks for your help. So a definite hessian is also semi definite? $\endgroup$ – Peter Pan Aug 8 '14 at 16:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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