Differentiating the composition of a scalar and a vector field using the chain rule

642 Views Asked by At

I have a vector $V=[v_0,v_1...v_n]'$ which is a function of another Vector $W=[w_0,w_1,...,w_n]'$. Here $'$ denotes the transpose.

The scalar field is $f=V'(W) V(W)$ where $V$ and $W$ are of the same dimension $(n+1), n=0,1,...,n$.

I need the gradient. So I think this is the chain rule and I need the Jacobian matrix - is this right?

$$del(f) = 2JV$$

where $J$ is the Jacobion representing the differentiation of the vector $V$ wrt the vector $W$.

Also I am confused about the form of the Jacobian. Standard form has each row as follows eg. row $0$ would be

$$[dv_0/dw_0, \, dv_0/dv_1, \, \cdots, \, dv_0/dv_n]$$

Now to prove this I set up a summation $$f=\sum_{i=0}^n v(w_0,w_1,...,w_n)(i)^2 $$

$$ \frac{df}{v(i)} = 2\sum_{i=0}^n v\cdot \frac{dv(j)}{di}, \quad j=0,1,...,n$$

Does this look right? - sorry the text may put my equation off skew