1
$\begingroup$

I have an cost function such as $$E(U)=\lambda \sum_{i=1}^{N} \int_{\Omega}r_iu_idx+\frac{1}{2\theta}\sum_{i=1}^{N} \int_{\Omega}(v_i-u_i)^2dx+ \frac{\beta}{2} \int_{\Omega}\left ( \sum_{i=1}^Nu_i-1\right )^2dx$$

subject to $0 \le u_i(x) \le 1$ for $i=1, ..., N$

The optimal condition is $$\frac {\partial E}{\partial u_i}=\lambda r_i+\frac{1}{\theta}(u_i-v_i)+\beta \left (\sum_{i=1}^Nu_i-1\right )=0$$

Now, from the above equation, I want to find the solution of $u_i$. Could you help me to prove that solution of $u_i$ will be $$u_i=v_i-\lambda \theta r_i - \frac {\beta \theta \left (\sum_{j=1}^N(v_j-\lambda \theta r_j)-1\right )}{1+N\beta \theta}$$

Because I have no idea why has it term $v_j-\lambda \theta r_j$ in the sum.

$\endgroup$
1
$\begingroup$

This is one method to obtain solution to system of linear equations $$\frac {\partial E}{\partial u_i}=\lambda r_i+\frac{1}{\theta}\left(u_i-v_i \right)+\beta \left (\sum_{j=1}^Nu_j-1\right )=0 $$

for all $i \in \{1, 2, ..., N\}$

Let $$t = \sum_{j = 1}^{N}u_j$$

We have $$\lambda r_i+\frac{1}{\theta}(u_i-v_i)+\beta t-\beta=0$$

Assuming $\theta \not = 0$ we get

$$u_i = \beta\theta - \beta\theta t + v_i - \lambda \theta r_i $$

So $$t = \sum_{j = 1}^{N}u_j = N\beta\theta - N\beta\theta t + \sum_{j = 1}^{N}v_j - \lambda \theta r_j $$

Assuming $1 + N \beta \theta \not = 0$ we get

$$t = \frac{ N\theta \beta + \sum_{j = 1}^{N}\left(v_j - \lambda \theta r_j\right)}{1 + N \beta \theta}$$

Finally we have $$u_i = v_i - \lambda \theta r_i + \theta \beta - \frac{\theta\beta\left( N\theta \beta + \sum_{j = 1}^{N} \left(v_j - \lambda \theta r_j\right) \right)}{1 + N \beta \theta} $$ $$ = v_i - \lambda \theta r_i - \frac{\theta \beta \left( \sum_{j = 1}^{N} \left(v_j - \lambda \theta r_j\right) - 1 \right)}{1 + N \beta \theta}$$

$\endgroup$
  • $\begingroup$ Perfect! +1. Thanks so much $\endgroup$ – John Jul 15 '15 at 7:15

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.