matrix
A[e1,e2,...,eN]

transformation

column i are image of standard basis ei

space(cols or rows)

a box

formed by columns or rows as edges

Each standard basis vector Ei is transformed to A column Ci

Av = AIv

one basis to another

column i are image of standard basis ei

orthogonal matrix vs unitary matrix

Projection Matrix

e = b - p
p = xa
a^t e = 0
a^t
(b - xa) = 0
x = a^t b / a^t a
p = xa

image

image

image

Proj-p = Pb
Pb = xa = (aTb/aTa)a
= (aaT/aTa)b
P = (aaT/aTa) = aaT/||a||
P = uuT (u = a/||a||)

image

image

Ax = b
b not in colspace(Ax) -> no solution
closest vector to b is b's projection on colspace(Ax)
get x when Ax = b-proj onto colspace(Ax)

b's projection onto a space is a vector
in the space which is closest to b

Inverse & Transpose

image

Rank

rank(A+B)≤rank(A)+rank(B)
rank(A+B)≥max(rank(A),rank(B))−min(rank(A),rank(B))

A MxN, B MxN
rank(A+B) <= min(M, N, rank(A)+rank(B))

Subset of rank one matrices is not subspace

b = p + e
b = Pb + e
e = b - Pb = (I-P)b
Pe = (I-Pp)
Pp = (I-Pe)