I need to minimize the following function: $f : \mathbb{R}^8 \to \mathbb{R}$, defined by \begin{align*}f(x_1, x_2, x_3, x_4, y_1, y_2, y_3, y_4)& =x_1x_2x_3x_4 + x_3 x_4 y_1 y_2 - x_2 x_3 y_1 y_4 + x_2 x_4 y_1 y_4 + x_1 x_3 y_2 y_4\\ &- x_1 x_4 y_2 y_4 + x_1 x_2 y_4^2 + y_1 y_2 y_4^2 \end{align*} with the constraint: $$x_1^2+x_2^2+x_3^2+x_4^2+y_1^2+y_2^2+y_3^2+y_4^2 \leq 1$$

I tried a few points which satisfy the constraint, and got a minimum of $\frac{-1}{16}$, but I am having trouble proving this rigorously.

I first tried letting Mathematica compute it but it would not terminate. Is there any quick way to do it rather than do calculus-style optimization in the interior or Lagrange multiplier on the boundary?

  • $\begingroup$ Thank you @Dr.MV, may I ask how you numerically approximated the maximum? What about the minimum? $\endgroup$ – Alvey Mar 16 '15 at 23:18

Hint: Note that the minimum must occur on the boundary, because if you get a negative result and you are not on the boundary, then you can scale the answer up by a constant factor and get on the boundary with a lower objective value, because your objective function is a homogeneous polynomial. So Lagrange multipliers are probably the way to go here, unless there's some trick hidden in the structure of the terms in the objective.


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.