In our calculus textbook we are given the following exercise:

Given function $f(x_1, x_2, \dots, x_n) = \sum_{i = 1}^{n}\sin(x_i) + \sin\left(\sum_{i=1}^{n}x_i\right)$ with $x_i \in (0, \pi)$ and $\sum_{i = 1}^{n}x_i \in (0, \pi)$ find it's extrema points and determine their type using only gradient and Hessian matrix.

Taking partial derivatives for two arbitrary variables $x_i$ and $x_j$ we get $$\partial_{i}f = \cos(x_i) + \cos\left(\sum_{i = 1}^{n} x_i\right) = 0,$$ $$\partial_{j}f = \cos(x_j) + \cos\left(\sum_{i = 1}^{n} x_i\right) = 0.$$ Subtracting them we get that $$\cos(x_i) = \cos(x_j),$$ and so all minima lie at point where all values will be the same. Now trying to find points of interest gives me the condition $$\cos(x_i) = \cos(nx_i).$$

From that point and on, I'm stuck. How can I proceed there?


You have all $x_i=x$ with $$ \cos x=-\cos nx $$ (note the minus sign that you have missed above) where $x,nx\in(0,\pi)$. The only solution to this equations is easy to see on the unit circle from symmetry (note $nx>x$) $$ nx=\pi-x. $$ Thus $x_i=x=\frac{\pi}{n+1}$.

Calculate the Hessian and verify that it is negative definite, hence the critical point is a local maximum.


Hint: According to your conclusion "all minima lie at point where all values will be the same", at mimimum point we have $(x_1, x_2, \dots, x_n) = (t, t, \dots, t)$ so to find $t \in (0 , \pi)$ you need to minimize the function

$$g(t) = n \sin(t) + \sin(nt)$$

Now go ahead and set derivative equal to zero...

By the way I assumed minimum exists !


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.