Perturbation Theory and $\delta J(v) = \langle DJ(v) \; , \; \delta v \rangle$?

97 Views Asked by At

In my optimization course, we are talking about using perturbation theory on bilinear functionals and stuff. We were told to use the following relationship to describe how the perturbation $\delta$ in the variable $v$ interacts with the perturbation in the functional $J$:

$$\delta J(v) = \langle DJ(v) \; , \; \delta v \rangle$$

So the perturbation of $J$ is equal to the inner product of the perturbation of $v$ and the differential of $J$ at $v$.

I can use this just fine, however I don't feel like I get where it comes from; we were just given this statement as a fact (or I missed the explanation).

How does this equivalence come about?

1

There are 1 best solutions below

0
On BEST ANSWER

One way to define the gradient $DJ$ of a vector function $J(v)$ is according to the approximation

$$J(v+\delta v)=J(v)+\langle DJ(v),\delta v\rangle + o(\delta v),$$ where $o(x)$ denotes a function that tends to $0$ as $x\to 0$.

The notation $\delta J(v)$ suggests a small change in the function $J(v)$ of size $\delta v$, that is, $$ \delta J(v)=J(v+\delta v)-J(v). $$ Combining these two equations and discarding the $o(\delta v)$ term (which is asymptotically negligible in magnitude compared to $\delta v$) yields the equation you wrote.