Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Figures
Orders:
A relation from A to B is a subset of the cartesian product R A B.
Thus the set of all relations is P(A B).
Inj(A; B)
Injections
&
BA
functions
Bij(A; B)
Bijections
&
P(A B)
relations
%
Surj(A; B)
Surjections
Example: Consider the set of matrices Mn (R) along with the order P D
where X P D Y if X Y is positive definite. (Mn (R), P D ) is a partial
order.
This order is used in the Gauss-Markov theorem.
1
MCS: f :
Lattice
(variables)
Loset
Poset
(parameters)
From Casella & Berger (2002) p. 78, exercise (2.15), (Betteley, 1977):
Let (X Y ) min(X, Y ) , and (X Y ) max(X, Y ).
FFTWE
Feasible Allocations
Weakly Pareto Efficient
Pareto Efficient
Walrasian Equilibrium
All Sequences X N
Bounded Sequences B(X), ` (X)
Cauchy Sequences C(X)
Convergent Sequences Conv(X)
Eventually Constant Sequences
s `p `q c0 c ` RN
Continuous functions C 0
Uniformly Continuous
Absolutely Continuous
Lipschitz Continuous
Bounded Derivative
We can take any non-empty set X and add algebraic structure, geometric structure, order structure and measurable structure. For example we
usually work with (R, h, iE , , , + , B(R) ), we use this structure
| {z }
| {z } |{z} |{z}
geometric order Algebraic measurable
Note that homeomorphism means that two spaces have the same topology,
isometry means two spaces have the same geometry.
Two measure spaces are measurably isomorphic if there exists a measurable bijection with a measurable inverse (CSZ: 7.6.17).
The Borel sigma algebra of any two polish spaces (complete and separable)
are measurably isomorphic iff the spaces have the same cardinality.
A function f : (X, dX ) (Y, dY ) is K-Lipschitz if K > 0 s.t.
(create diagram)
Uniformly Continuous
Lipschitz
Bounded Derivative
C0
C1
..
.
Smooth C
10
Topology:
f : (X, X ) (Y, Y ), it is easier for f to be continuous with a finer
domain or a coarser co-domain.
0 then it will remain continuous with the finer
If f is continuous and X X
0
topology X .
If f is continuous and Y0 Y then it will remain continuous with the
coarser topology Y0 .
Whenever X = P(X) is the topology on the domain, every function on this
domain will be continuous.
Whenever Y = {, Y } is the topology on the co-domain, every function
into this co-domain will be continuous.
For a function f : (X, {, X}) (Y, P(Y )) the only continuous functions
are the constant functions.
Given a non-empty set X, let TOP(X) be the collection of all topologies on
X
X. Since TOP(X) 22 , we have that whenever X is finite TOP(X) will
be finite as well.
%
{, X}
&
coarsest topology
&
2X
finest topology
Using the -order, the indiscrete topology {, X} will always be the -least
set in (TOP(X), ), and the discrete topology 2X will be the -greatest set.
G.O. (I believe (TOP(X), ) is a complete lattice because the intersection
of topologies is a topology.)
In the indiscrete topology {, X}: every sequence converges, every set is
connected, every set is compact.
In the discrete topology 2X only eventually constant sequences converge, it
is difficult for a set to be compact or connected, only finite sets are compact.
Given a non-empty set X consider two topologies (X, 1 ) and (X, 2 ) such
that 1 2 . Any sequence that converges in 2 must converge in 1 , any
set that is compact in 2 must be compact in 1 , any set that is connected
in 2 must be connected in 1 .
11
Convexity
Let V be a real vector space and let S be any non-empty subset, then:
Aff(S)
Affine Hull
&
span(S)
Conv(S)
Linear Hull
Convex Hull
&
%
Cone(S)
Convex Conic Hull
For a canonical exercise let the vector space be R2 and let S = {e1 , e2 }.
We have: span(S) = R2 .
Aff(S) = {(x1 , 1 x1 ) : x1 R}
Cone(S) = R2+ (the positive orthant)
Conv(S) = 1 (the 1-simplex of probabilities)
CREATE this figure in Mathematica
Note if our set is a singleton (consists of only one vector) S = {x} then:
span(S) is a line, Cone(S) is a ray, and Conv(S) = Aff(S) = {x}.
Note that if S is a vector subspace of V it is equal to the intersection of all
vector subspaces containing it. If S is a subset of a vector space V , then the
smallest vector subspace that contains S is span(S) which is equal to the
intersection of all vector subspaces of V that contain S. The same holds for
affine subspaces, conic hulls and convex hulls.
Datorro Convex book p. 65: Conv(S) Aff(S) = Aff(S) = Aff(S) =
Aff Conv(S).
A set of vectors in Rn are affinely independent if there is no proper affine
subspace which contains all of them.
To visualize affine independence consider the following three affinely independent
R2 , along
vectors
in
with the affine hulls containing any two of
0
2
1
them:
,
,
2
0
2
12
6
Aff
0 2
,
2 0
4
Aff
2
-4
-2
1 0
,
2 2
-2
-4
Aff
1 2
,
2 0
-6
Critical points of f
Local extrema
Global Extrema
13
Case 1 2
1
+ 0 0
2
0 + 0
3
0 0 +
4
+ + 0
5
+ 0 +
6
0 + +
7
+ + +
8
0 0 0
The following diagram illustrates the points these cases correspond to when
p1 , p2 , w > 0:
2.0
1.5
1.0
0.5
6
-2
5
1
-1
-0.5
3
-1.0
Notice that case 7 is impossible because if x1 = x2 = 0 then the first constraint cannot possibly bind. The only time case 7 would hold is if the
constraint passed the origin.
Note: if we know that the objective function u(x1 , x2 ) is increasing in both
of its arguments, then the only relevant cases are: 1, 4, and 5.
This diagram was inspired by Nolan Millers handout: The Kuhn Tucker
Conditions and you.
The following diagram illustrates the relationship between types of concavity:
14
strictly quasi-concave
%
&
quasi concave
strictly concave
&
%
concave
A constant function is a function that is concave but not strictly quasiconcave. Are there other functions with this property?
An increasing transformation of a concave function is quasi-concave. (Similarly: an increasing transformation of a homogeneous function is homothetic,
and an increasing transformation of an spm function is q-spm.)
The standard normal distribution is an example of a function that is quasiconcave but not concave:
15
0.4
0.3
0.2
0.1
-10
-5
10
&
quasi concave
strictly concave
&
concave
pseudo concave
16
Measure
Given a non-empty set X the collection of all -algebra on X is a complete lattice:
%
{, X}
&
coarsest algebra
&
2X
finest algebra
&
algebra
%
algebra
Also look at Border p. 499 for a table of normed Reisz spaces and their
duals.
Equivalently we have the following nested sequence of boxes:
-algebra
algebra
ring
semi-ring
17
system
%
algebra
&
system
algebra
(A)
%
&
(A)
(A)
&
%
f (A)
18
19
20
Abstract Algebra:
Magma (Binary Algebraic Structure) closed under operation
Semigroup (Associative)
Monoid (Identity)
Group (Inverse)
Abelian Group (Commutative)
Ring, Field, Vector Space over a Field (Linear Algebra in a Nutshell), Algebra (just a VS with vector multiplication)
Linear Algebra:
Tensor Product
Kronecker Product
Outer Product
Operations on sets
Given X 6= , intersection, unions and symmetric difference are binary algebraic structures.
: P(X) P(X) P(X), the intersection operation is commutative, associative, and has an identity X, however it is not invertible.
The union operation is commutative, associative, and has an identity ,
however it is not invertible.
(P(X), ) is an Abelian group (like addition), the identity is , and every
21
22