Let $y=f(x)=(x_1^2+2x_2, x_1x_2-3x_1)$

Is the linear approximation just $f(y)=f(x)+A(y-x)$ whenever $y$ is approximately near $x$?

I know that if I calculate the Jacobian matrix, I can get that $Df=\begin{pmatrix} 2x_1 & 2\\ x_2 & 0 \end{pmatrix}\implies Df(1,1)=\begin{pmatrix} 2 & -2\\ 1 & 0 \end{pmatrix}$.

I'm not sure how to proceed from here.

  • 2
    $\begingroup$ Your $Df$ is wrong $\endgroup$ – user251257 Oct 9 '15 at 7:13

A differentiable function $f:\mathbb{R}^2\to\mathbb{R}^2$ can be approximated by it's derivative in the sense that given a point, say, $(1,1)$, you can write:

$$f(1+\delta x,1+\delta y)\approx f(1,1)+Df(1,1)(\delta x,\delta y) $$

where $Df(1,1)$ is a matrix applied to the vector $(\delta x,\delta y)$. So, yes, if $y$ is near $x$ (as $(1+\delta x,1+\delta y)$ is near $(1,1)$), then you can approximate the difference $f(y)-f(x)$ by the derivative evaluated at $x$, applied to the difference $y-x$. That is:

$$f(y)\approx f(x)+Df(x)(y-x)$$

Compare this to the one dimensional case, where the tangent line to the graph of a differentiable function locally approximates the graph, in the sense that:

$$f(y)\approx f(x)+f'(x)(y-x)$$


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.