Thursday
finish last lecture
V - finite dimension vector space / F
V' - dual space L (V,F)
- set of linear maps; can think of as functions on V
- think about pairing
- V' x V goes to F
- can reverse the roles
- given functional phi
< >
duality between V and V'
a new layer is created in linear algebra
a canonical isomorphism exists between two things
rare
highly prized
This is one of them
Any vector v can be treated as an element of the dual space or
Canonical map
Theorem: this map D is an isomorphism
the dual of the dual
Proof:
sufficient to show that D is injective, which is equivalent to saying that the null space
---
contrast that to
Any two vector spaces of dimension n are isomorphic
special case in F^n where $\exists$ a canonical basis
powerful statement: isomorphism without defining bases
all vector spaces come in pairs
Dual maps
---
numerical representation
if choose basis of V
dual basis
convenient to think of the pairing of the duals as rows on the left, column on right
- so the product expresses the relationship as a multiplication
- comment on physics notation using b^i to indicate summation over the product: b^i a_i
- discussion of duality, reversing the map
Use pairing for proof
< $\phi$ | T | V >
means applying T to V
or other way
---
Application
36:00
zoo of vector spaces
all VS of dimension n have dual
- Lagrangian interpolation or polynomial interpolation
- use P
- dimension n + 1
- define Lagrange polynomials
- expression of product
$\large\pi_{i\ne j}$ $\frac{t-a_i}{a_a-a_a}$
- check that these polynomials have this property
- p sub i of a sub j is the Kroneker delta
- Theorem: just knowing linear algebra
- notion of dual basis; functionals form basis of V prime
- value at point b is always given by combination of values at the points at a sub i with universal coefficients c sub i
- for all polynomials
- if know polynomial at n+1 points, know its value anywhere
- know value of any polynomial anywhere given values at a sub i
there is a universal formula
cool application
One more thing for Dual spaces
- linear maps
What about subspaces?
There must be a subspace in V'
subset of all phi in V' such that
U ^0 is defined
Example
V is
lambda phi a = valuation at point a. where lambda is in R
U zero is all polynomials whose value at a is zero
reference to homework problems where had to find values of polynomials
not only dimension but
define the null space of U
dimension is less by 1;
term is co-dimension 1
Next result: there is a pattern where dimensions of U and U zero are complementary to each other
U zero is subspace, not just subset of V'
dimension is = to dimension
3.125
---
51:14 picture is complete:
if T is injective, then null space is zero.
annihilator is V'
T is injective iff T' is surjective
---
Eigenvectors and eigenvalues start now, follow after the exam
Preliminaries
Short Chapter 4 about polynomials: complex numbers
new notation for polynomials
Definition:
4.6
4.8
Chapter 4 has proof of theorem that relies on complex analysis
not responsible for proof, only result
---
eigenvector and eigen value
Chapter 5
T is an operator ; still a linear map, but from V to V
Example: rotation
Suppose you have an operator
Two things
Subspace U inside V is called invariant under T, or T-invariant
if
Lemma:
suppose you have
vi, .... , vk are eigenvectors with eigenvalues that are distinct
then this set {v1, ... vk} is linear independent
all info on midterm is on bcourses
Tuesday review here; exam Thursday