Ex. 1

Find the minimum of \(f(x,y) = x^2 + xy + y^2 in (x, y) \in \mathbb{R}^2\)

Solution:

\(f_x(x,y) = \frac{\partial}{\partial x} (x^2 + xy + y^2)\)

\(f_y(x,y) = \frac{\partial}{\partial y} (x^2 + xy + y^2)\)

Now the stationary conditions are

\(\frac{\partial f}{\partial x} = 0\) and \(\frac{\partial f}{\partial y} = 0\)

That means

\(2x + y = 0\)

\(x + 2y = 0\)

From second equation above we get x = -2y

replacing x in first equation gives

\(2 * (-2y) + y = 0 => y = 0\)

putting y = 0 in any of the above equations, we get x = 0

Next is to compute all 4 second partial derivatives.

\(f_{xx}(x,y) = \frac{\partial}{\partial x} (2x + y)\) = 2

\(f_{xy}(x,y) = \frac{\partial}{\partial y} (2x + y)\) = 1

\(f_{yx}(x,y) = \frac{\partial}{\partial y} (x + 2y)\) = 1

\(f_{yy}(x,y) = \frac{\partial}{\partial y} (x + 2y)\) = 2

Hessian Matrix

\(H = \begin{pmatrix} f_{xx} & f_{xy} \\f_{yx} & f_{yy} \end{pmatrix}\) = \(\begin{pmatrix} 2 & 1 \\1& 2 \end{pmatrix}\)

Thus H is positive and point (0,0) is minimum.

Ex. 2

For \(f(x)= x^{4}\) in \(\mathbb{R}\), it has a global minimum at x = 0. Find its new minimum if a constraint \(x^{2} \geq 1\) is added.

Solution:

Using Penalty method

\(f(x)= x^{4}\) and \(g(x)=x^{2}-1\)

Now using a penalty parameter \(\mu\) we have

\(\amalg = f(x) + \mu [g(x)]^2\) = \(x^{4} + \mu (x^{2}-1)^2\) = \(x^{4} + \mu (x^{4}+1-2x^{2})\)

From \(\amalg'(x) = 0\), we have

\(4 x^{3} + 4 \mu x^{3} -4 \mu x= 0\)

on simplifying we get

\(x^{2} = \frac {\mu}{(1 + \mu )}\)

for very large value of \(\mu\), we will have

\(x^{2} = \frac {\mu}{(1 + \mu )} \approx \frac {\mu}{\mu } \approx 1\)

so x can be 1 or -1 so the new minimum will be at x = 1

Ex. 3

Use a Lagrange multiplier to solve the optimization problem \(min f(x,y) = x^{2} + 2xy + y^{2}\), subject to \(y = x^{2} - 2\)

Solution:

\(f(x,y) = x^{2} + 2xy + y^{2}\)

\(h(x, y) = x^{2} - y -2\)

\(\phi = f(x, y) + \lambda h(x,y)\)

\(\phi = (x^{2} + 2xy + y^{2}) + \lambda (x^{2} - y -2)\)

\(\frac{\partial \phi}{\partial x} = 2x + 2y + 2\lambda x = 0\)

\(\frac{\partial \phi}{\partial y} = 2x + 2y - \lambda = 0\)

\(\frac{\partial \phi}{\partial \lambda } = x^{2} - y -2= 0\)

From the second equation, \(2x + 2y - \lambda = 0\) => \(\lambda = 2(x + y)\)

Having \(\lambda = 2(x + y)\) in the first equation gives us

\(2x + 2y + 2(2(x + y)) x = 0\)

\(2x + 2y + 4x^2 + 4xy = 0\)

\(2(x + y + 2x^2 + 2xy) = 0\) which gives us \(y = \frac{-2x^2 - x}{2x + 1}\)

Having \(y = \frac{-2x^2 - x}{2x + 1}\) into the third equation gives us

\(x^{2} - 2 - (\frac{-2x^2 - x}{2x + 1}) = 0\) which after simplifying gives us x = 1 or x = - 2.

Now with x = 1, the third equation gives us y = -1 and with in x = -2 gives us y = 2.

Finally we have two points (1, -1) and (-2, 2).

The optimality for \(min f(x,y) = x^{2} + 2xy + y^{2}\) is (1, -1) with \(f_{min} = 0\) and (-2, 2) with \(f_{min} = 0\)

Resources:

https://www.khanacademy.org/math/multivariable-calculus/applications-of-multivariable-derivatives/lagrange-multipliers-and-constrained-optimization/v/constrained-optimization-introduction