1/47
Looks like no tags are added yet.
Basis Criteria
list of vectors satisfies 2 of the 3:
linearly independent set
spans V
number of elements = dim(V)
Reduce a spanning set to a basis
dim(span(v1,…,vk)) = # pivots in REF of [v1…vk]
A linear map T:V→W is an isomorphism iff
- (v1,…vk) is a basis of V and (Tv1,…,Tvk) is a basis for W
-dim(V) = dim(W)
Dimension of ker(A)
# of free variables in A
rank(T)
= dim(Im(T))
= # of pivots in A of T
nullity(T)
= dim(ker(T))
= # of free variables
rank - nullity theorem
rank(T) + nullity (T) = dim(V)
length (norm) of a vector
sqrt(v*v)
proj_w(v)
w*<v,w>/||w||²
Inner product properties
distributive, homogeneous, non-negative
symmetric: <v,w> = conjugate(<w,v>)
definite: if <v,v> = 0, v=0
orthagonality
<v,w>=0 and v and w are non-zero
an orthogonal list of non-zero vectors is
linearly independent
pythagorean theorem
||v1+…+vn||²=||v1||²+…+||vn||²
cauchy-schwarz theorem
|<v,w>|<=||v|| ||w||
equal is v=cw
triangle inequality
||v+w||<= ||v|| + ||w||
orthonormal basis
(vi,…,vn) is orthogonal for i~=j, but <vi,vi> =1
Gram-Schmidt Process
start with basis (v1,…,vn)
u1=v1/||v1||
~uj=vj - Σ<vj,uk>uk
uj=~uj/||~uj||
orthogonal complement of u (subspace of V)
u^perp = {<u,v>=0}
ex. u=x-y plane, complement is z axis
Pu(V)
Pu(V)=Σ<v,ui>ui
if u’s are an ON basis
Change of Basis matrix B1→B2
solve [B2|ai] where ai is cols of B1
[T]B1B2
columns of T(vi) in terms of B2 (where vi are col of B1)
[T]B1B2 to [T]B1’B2’
[T]B1’B2’ =SAS-1
S is cob matrix from B1 to B1’
if given an eigenbasis, S is ELs and A is EVs diagonally
[Pu]std
=QQT where Q=[u1…uk] where u is ON basis for U
=A(ATA)-1AT where A is matrix of a non ON basisA
A matrix can be diagonalized iff
it has an eigenbasis
Similar matrices
A=SBS-1 and B=SAS-1
A and B have the same rank and kernel
an upper triangular matrix is invertible iff
not all diagonal entries are 0p
parallelogram equality
||v+w||² + ||v-w||²=2||v||² + 2||w||²
L1 norm
||v||1= Σ|xi|
L_infinity norm
||v||inf=max|xi|
L_2 norm on continuous functions
0 to 1 integral of f²
operator norm
||T||op= max ||Tv|| for ||v||=1
Isometry
T: V→W is an isometry if
T is surjective
and ||Tv||=||v||
ex. rotation or reflection
Isometry T:V→W has (u1,…,un) ON basis of V
(Tu1,…,Tun) is an ON basis of W
A=[T]std where T is an isometry
I=ATAO
Orthogonal Matrix
A-1=AT
represent isometries
Unitary Matrix
A-1=A*
QR Decomposition
A=QR where Q is orthogonal, and R is upper triangular
Q is gram schmidt performed on the columns of A
R = Q-1A = QTA
Basis for U perp
if U is a subspace of Fn with ON basis (u1,…uk),
basis for U perp is (uk+1,…un)pro
proj onto a basis
sum of the projection onto each element of the basis
Show that if U and W are subspaces of a finite-dimensional vector space V , then
dim(U+W) <= dim(U) + dim(W)
show the bases (u1-n,w1-m) span U+W by showing all x in U+W are reached by it.
(u1-n,w1-m) = n + m elts (dimU +dimW), it can only be reduced to be made a basis
theorem 4.9 (change of basis with inner products)
if B1 and B2 are orthonormal bases, [v]B1=[<v,b11>,…,<v,b1n>]
change of basis from standard to another basis
if basis has matrix U, cob matrix is U-1
if the basis is also ON, cob matrix is UT
theorem 3.66 (every T over finite dimensional space has at least one eigenvalue)
for n=dim(V)
the set {v,Tv,…Tnv} has n+1 elts, so lin dep, set lin combo = 0
factor like a polynomial: (T-c1I)…(T-cmI)v=0
one factor has to be 0, so at least one is an eigenvalue
prove an orthogonal list is linearly independent
assume by contradiction some vi is dependent
apply <_,vi> to both sides
prove the pythagorean theorem
expand using inner products and summations
prove cauchy-schwarz theorem
v=proj_w(v)+u where u is perpendicular to w
apply pythagoras (square norm) and evaluate
prove the triangle inequality
square both sides
use the identity that <w,v> is the conjugate of <v,w>
use cauchy schwarz
eigenvalues of a projection map
only 0 and 1