For any $x, y$ in real, find the minimum value of $9x^2+2y^2-8xy-6x+11$
-
1You can compute the partial derivatives – Dr. Sonnhard Graubner Jan 31 '19 at 15:35
-
2In the future please take the time to enter your question as text instead of pasting in a picture. Images are neither searchable nor accessible to screen readers, nor do they show up in summaries. – amd Feb 01 '19 at 15:54
7 Answers
Let $$9x^2+2y^2-8xy-6x+11=k$$ Thus, the the following quadratic equation of $y$ $$2y^2-8xy+9x^2-16x+11-k=0$$ has real solutions, which says $$16x^2-2(9x^2-6x+11-k)\geq0$$ or $$(x-3)^2\leq k-2,$$ which gives $$k\geq2.$$ The equality occurs for $x=3$ and $y=6,$ which says that $2$ a minimal value.
- 194,933
You can use some concepts from multivariable calculus, namely looking for where the gradient vector of the function is equal to $\langle 0,0\rangle$. (This does not return a minimum, but rather, critical points). $$\nabla f(x,y) = \langle18x -8y - 6, 4y - 8x\rangle$$ Solving for $18x - 8y - 6 =0$ and $4y -8x =0$, we find that a critical point exists at $(3,6)$. Plugging this in, we find that the minimum of the function is $f(3,6) = 2$.
- 899
Finding the critical point of the quadric by computing partial derivatives, as suggested in other answers, is pretty quick and easy, but you can always rewrite the quadric as a linear combination of squared terms by making the ansatz suggested here. In fact, there’s a purely mechanical algorithm akin to Gaussian elimination, which you can read about here.
If you start off with an integer matrix, this algorithm uses only integers, which is handy when you’re calculating by hand, but they can get quite large. I’m willing to work with rational numbers, so I’ll use an equivalent algorithm that uses the operation $R_i\to R_i-{a_{i1}\over a_{11}}R_1$ to clear the column below a pivot, and similarly for the column operations. Omitting the gory details, this algorithm starts with $$\left[\begin{array}{c|c}A&I\end{array}\right] = \left[\begin{array}{ccc|ccc}9&-4&-3 & 1&0&0 \\ -4&2&0 & 0&1&0 \\ -3&0&11 & 0&0&1\end{array}\right]$$ and produces $$\left[\begin{array}{c|c}D&P^T\end{array}\right]=\left[\begin{array}{ccc|ccc}9&0&0 & 1&0&0 \\ 0&\frac29&0 & \frac49&1&0 \\ 0&0&2 & 3&6&1 \end{array}\right].$$ We can already see from this that the minimum value is $2$, but continuing on, we have $D=P^TAP$, but want $A=P^{-T}DP^{-1}$, so we need to invert $P$. This can be done by applying the inverses of the column operations that were used to produce $P$, reversed, to the identity matrix. (This is just a variation of the well-known algorithm for inverting a matrix via Gaussian elimination.) This results in $$P^{-1} = \begin{bmatrix}1&-\frac49&-\frac13\\0&1&-6\\0&0&1\end{bmatrix},$$ so that $$9x^2-8xy+2y^2-6x+11 = 9\left(x-\frac49y-\frac13\right)^2+\frac29\left(y-6\right)^2+2(1)^2.$$
- 53,693
Any extreme values of $f(x,y)$ must occur at critical points, that is, at points where $f_x = f_y = 0$. In this example, taking these derivatives gives you a system of linear equations.
- 4,997
You would need to solve the system of linear equations given by partial derivates
$$\left\{ \begin{array}{l} \frac{\partial}{\partial x}f(x,y)=0\\ \frac{\partial}{\partial y}f(x,y)=0 \end{array}\right. \Rightarrow \left\{ \begin{array}{l} 18x-8y-6=0 \\ 4y-8x=0 \end{array}\right. \Rightarrow \left\{ \begin{array}{l} y=2x \\ y=6 \end{array}\right. $$
- 4,685
Define a function $F(x,y) = 9x^2 + 2y^2 - 8xy - 6x +11$
Then according to the Lagrange Multiplier Method, we know that there is no restriction, so we can define function $G(x,y)=F(x,y)+0\lambda=F(x,y)$,and then we take the partial derivatives of $x$ and $y$:
$$G'x(x, y) = 18x - 8y - 6 = 0$$
$$G'y(x, y) = 4y -8x = 0$$
Solve the equations, so we discover
$$x = 3, y = 6$$
Calculate it, it's the minimum value 2.
- 13
-
1I do not think that this is related to the Lagrange Multiplier Method – Claude Leibovici Jan 31 '19 at 16:33
-
Well, just this question has no restrictions, no more other functions, and can get derivations directly, it is similar to Lagrange in some way. – ItoMakoto Feb 01 '19 at 07:33
-
2I agree with @Claude Leibovici : for deserving the name "Lagrange multiplier", you need two functions $F$ and $G$. It is like saying orange color is similar to yellow color or to red color ; they are neighbor colors but definitely not similar. Otherwise, you would have to consider, by transitivity, that yellow is similar to red :) – Jean Marie Feb 01 '19 at 15:41
-
1You don’t even have a Lagrange multiplier in your answer. Why even mention the method when all you’re doing is computing critical points of $F$? – amd Feb 01 '19 at 15:56
-
So you mean I should define a function G which equals to F because with no restriction and λ is meaningless, or other things I may have misconception?@Jean Marie – ItoMakoto Feb 01 '19 at 16:23
-
1The bottom line is that this is not a problem for which the Lagrange multiplier method is suitable precisely because there’s no constraint. – amd Feb 02 '19 at 22:41
Among several methods, a general one to find the extremums of a function $f(x,y)$ consists in solving the system : $$\begin{cases} \frac{\partial f}{\partial x}=0 \\ \frac{\partial f}{\partial y}=0 \end{cases}$$ This leads to one or several points $(x_k,y_k)$ . For each one compute the corresponding value of $f(x_k,y_k)$. Comparing them (if they are several) allows to see if each point corresponds to a maximum or a minimum of $f(x,y)$ , local or general.
Several extremums can be observed for example in case of some polynomials of degree >2.
In the case of $$f(x,y)=9x^2+2y^2-8xy-6x+11$$ $$\begin{cases} \frac{\partial f}{\partial x}=18x-8y-6=0 \\ \frac{\partial f}{\partial y}=4y-8x=0 \end{cases}$$ Solving is very easy and gives : $x=3$ and $y=6$. $$f(3,6)=9(3^2)+2(6^2)-8(3)(6)-6(3)+11$$ $$f(3,6)=2$$ There is only one extremum. Another point (any one), for example $f(0,0)=11>f(3,6)$ . Thus the extremum is a minimum and the minimum value of the function is $2$.
- 66,221
- 3
- 37
- 87
