4
$\begingroup$

I want to minimize a quadratic equation, but Mathematica does not solve it and only rewrites the solution. I wonder if there exist any way I can skip the computation by hand (I can solve it using Lagrangian method, but it is too long).

y = {y1, y2, y3}
yp = {yp1, yp2, yp3}
ypp = {ypp1, ypp2, ypp3}
Minimize[{(ypp1 - yp1)^2 + (ypp2 - yp2)^2 + (ypp3 - yp3)^2,  
           Mean[ypp] == 0 && Variance[ypp] == 1}, ypp]
$\endgroup$
5
  • $\begingroup$ I'd shift ypp and work out the new mean and variance, then use Minimize[{Total[(ypp)^2], Mean[ypp] == a, Simplify[Variance[ypp], Assumptions -> {ypp \[Element] Reals}] == b}, ypp, Reals]. $\endgroup$ May 26 '13 at 16:27
  • $\begingroup$ This is a repeat of this question: mathematica.stackexchange.com/questions/25845/… (last version in the answer) $\endgroup$
    – bill s
    May 26 '13 at 16:30
  • $\begingroup$ @bills You can vote to mark it as a duplicate. To do that (in case you don't know), click the "close" link > duplicate > enter the URL of the question. $\endgroup$
    – rm -rf
    May 26 '13 at 16:38
  • $\begingroup$ But the program above produces: Minimize::infeas: There are no values of {ypp1,ypp2,ypp3} for which the constraints 1/3 (ypp1+ypp2+ypp3)==0&&1/3 (ypp1^2+ypp2^2-ypp2 ypp3+ypp3^2-ypp1 (ypp2+ypp3))=={5,3,6,3} are satisfied and the objective function ypp1^2+ypp2^2+ypp3^2 is real valued. $\endgroup$
    – remo
    May 26 '13 at 16:42
  • $\begingroup$ @bill I believe there is a difference. yp1, yp2, and yp3 are symbolical here, contrary to v in your previous answer, and you don't get a result. $\endgroup$ May 26 '13 at 18:59
7
$\begingroup$

Firstly, variance uses Conjugate and that stumbles Minimize if ypp is symbolic. It can be mended with ComplexExpand. Secondly, evaluation is going for long. I supect it is due to many variants of relative positions of yp components. Setting the order explicitely, say yp1<yp2<yp3, gives the result:

y = {y1, y2, y3}
yp = {yp1, yp2, yp3}
ypp = {ypp1, ypp2, ypp3}
Minimize[{(ypp1 - yp1)^2 + (ypp2 - yp2)^2 + (ypp3 - yp3)^2, 
  Mean[ypp] == 0 && Variance[ypp] == 1 // ComplexExpand,  yp1 < yp2 < yp3}, ypp][[1, 1, 1, 1]]

$$ -\frac{4 \sqrt{\text{yp1}^2-\text{yp1} \text{yp2}-\text{yp1} \text{yp3}+\text{yp2}^2-\text{yp2} \text{yp3}+\text{yp3}^2}}{\sqrt{3}}+\text{yp1}^2+\text{yp2}^2+\text{yp3}^2+2 $$

$\endgroup$
2
  • $\begingroup$ Thanks a lot. It's really wonderful why such a simple question takes too long. $\endgroup$
    – remo
    May 26 '13 at 17:14
  • 1
    $\begingroup$ Nice trick. (+1) $\endgroup$ May 26 '13 at 19:01

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.