Representing inner product between a Matrix and column vector, pseudo inverse, and projection

214 Views Asked by At

The inner product of two column vectors v and w is defined as:

$$v \cdot w = w \cdot v = <v,w>=v^T w = <w,v>= w^T v =|proj_w(v)| = |proj_v(w)| = \text{a scalar value}$$

Now if I were to generalize this concept of Inner product between a Matrix A and column vector x, What do we call this type of inner product? and how do we define the order of the matrix in the inner product operator $<A,x>$ verses $<A,x>$...

do we write $A^T x$ as $<A,x>$

... or do we write $A^T x$ as $<x, A>$?

Does Projection of A onto x equal:

$$|proj_x(A)| = <A,x> / <A,A>$$

or does it equal:

$$|proj_x(A)| = <x,A> / <A,A>$$

Given Pseduo inverse:

$$x=Ay$$ $$A^Tx=A^TAy$$ $$y=\frac{A^Tx}{A^TA}$$

$$A^{\dagger} = \frac{A^T}{A^TA}$$

$$y=A^{\dagger}x$$

thus, pseudo inverse corresponds to an orthogonal projection of vector x onto the manifold of A...

can we say:

$$A^{\dagger}x = proj_x(A)$$?

or would that be:

$$A^{\dagger}x = proj_A(x)$$?