Static Optimization problem assumptions

49 Views Asked by At

Consider the following Constrained Optimization Problem:

$$max: -2(x+y)^2 +5x-y$$ Subject to: $$x≤B$$ $$-x+2y≤3$$ $$y≥0$$ with$$ \beta = 0$$ I have found the solution that solves the problem:$$x=\beta$$ $$y=0$$ $$\lambda_1 = 5-4\beta$$ $$\lambda_2 =0$$ $$\lambda_3=4\beta +1$$I found several other sets of solutions by not making any prior assumptions about the constraints, i.e., whether or not they were binding or slack. I can see that the original function is decreasing in positive values of $y$, would it be reasonable to assume $y=0$ and then solve the problem? Or is working the entire problem out the only way to be sure that the solution is correct?

1

There are 1 best solutions below

2
On

Solving with the help of Lagrange Multipliers and introducing some slack variables $(\epsilon_i)$ to reduce the inequalities to equalities we have with $f(x,y) = -2(x+y)^2+5x-y$

$$ L(x,y,\lambda,\epsilon) = f(x,y)+\lambda_1(x-\beta+\epsilon_1^2)+\lambda_2(-x+2y-3-\epsilon_2^2)+\lambda_3(y-\epsilon_3^2) $$

The stationary conditions are

$$ L_x = 5+\lambda_1-\lambda_2-4(x+y)=0\\ L_y = -1+2\lambda_2+\lambda_3-4(x+y) = 0\\ L_{\lambda_1} = -\beta+x+\epsilon_1^2 = 0\\ L_{\lambda_2} = 2y-3-x-\epsilon_2^2 = 0\\ L_{\lambda_3} = y-\epsilon_3^2 = 0\\ L_{\epsilon_1} = \lambda_1\epsilon_1 = 0\\ L_{\epsilon_2} = \lambda_2\epsilon_2 = 0\\ L_{\epsilon_3} = \lambda_3\epsilon_3 = 0 $$

Solving those equations we have the possible stationary points as well as the $f(x,y)$ value according to the table

$$ \begin{array}{ccccccccc} x & y & \lambda_1 & \lambda_2 &\lambda_3 &\epsilon_1 &\epsilon_2 &\epsilon_3 & f\\ \beta & \frac{\beta +3}{2} & 9 \beta +\frac{9}{2} & 3 \beta +\frac{7}{2} & 0 & 0 & 0 & -\frac{\sqrt{\beta +3}}{\sqrt{2}} & -\frac{3}{2} (3 \beta (\beta +1)+4) \\ \beta & \frac{\beta +3}{2} & 9 \beta +\frac{9}{2} & 3 \beta +\frac{7}{2} & 0 & 0 & 0 & \frac{\sqrt{\beta +3}}{\sqrt{2}} & -\frac{3}{2} (3 \beta (\beta +1)+4) \\ \beta & 0 & 4 \beta -5 & 0 & 4 \beta +1 & 0 & -\sqrt{-\beta -3} & 0 & (5-2 \beta ) \beta \\ \beta & 0 & 4 \beta -5 & 0 & 4 \beta +1 & 0 & \sqrt{-\beta -3} & 0 & (5-2 \beta ) \beta \\ \beta & -\beta -\frac{1}{4} & -6 & 0 & 0 & 0 & -\sqrt{-3 \beta -\frac{7}{2}} & -\frac{1}{2} \sqrt{-4 \beta -1} & 6 \beta +\frac{1}{8} \\ \beta & -\beta -\frac{1}{4} & -6 & 0 & 0 & 0 & -\sqrt{-3 \beta -\frac{7}{2}} & \frac{1}{2} \sqrt{-4 \beta -1} & 6 \beta +\frac{1}{8} \\ \beta & -\beta -\frac{1}{4} & -6 & 0 & 0 & 0 & \sqrt{-3 \beta -\frac{7}{2}} & -\frac{1}{2} \sqrt{-4 \beta -1} & 6 \beta +\frac{1}{8} \\ \beta & -\beta -\frac{1}{4} & -6 & 0 & 0 & 0 & \sqrt{-3 \beta -\frac{7}{2}} & \frac{1}{2} \sqrt{-4 \beta -1} & 6 \beta +\frac{1}{8} \\ -3 & 0 & 0 & 17 & -45 & -\sqrt{\beta +3} & 0 & 0 & -33 \\ -3 & 0 & 0 & 17 & -45 & \sqrt{\beta +3} & 0 & 0 & -33 \\ -\frac{1}{2} & \frac{5}{4} & 0 & 2 & 0 & -\sqrt{\beta +\frac{1}{2}} & 0 & -\frac{\sqrt{5}}{2} & -\frac{39}{8} \\ -\frac{1}{2} & \frac{5}{4} & 0 & 2 & 0 & -\sqrt{\beta +\frac{1}{2}} & 0 & \frac{\sqrt{5}}{2} & -\frac{39}{8} \\ -\frac{1}{2} & \frac{5}{4} & 0 & 2 & 0 & \sqrt{\beta +\frac{1}{2}} & 0 & -\frac{\sqrt{5}}{2} & -\frac{39}{8} \\ -\frac{1}{2} & \frac{5}{4} & 0 & 2 & 0 & \sqrt{\beta +\frac{1}{2}} & 0 & \frac{\sqrt{5}}{2} & -\frac{39}{8} \\ \end{array} $$

NOTES

1-Some values are duplicate due to the adoption of $\epsilon_i$ squared.

2-Solutions with at least one $\epsilon_i = 0$ are located at the boundary. As we can observe, all solutions are at the feasible region boundary.

3-Once the $\beta$ value is fixed, the maximum and minimum can be chosen.