Scribd
Upload
10 views50 pages

Classical Programming_Lecture 1

The document discusses classical programming, focusing on maximizing or minimizing a function subject to constraints, and introduces the unconstrained case where no constraints are present. It outlines necessary conditions for local maxima, including first and second order conditions, and provides examples to illustrate these concepts. Additionally, it touches on the application of these principles in economic institutions such as households, firms, and governments.

Uploaded by

Beatrice Martyan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
10 views50 pages

Classical Programming_Lecture 1

The document discusses classical programming, focusing on maximizing or minimizing a function subject to constraints, and introduces the unconstrained case where no constraints are present. It outlines necessary conditions for local maxima, including first and second order conditions, and provides examples to illustrate these concepts. Additionally, it touches on the application of these principles in economic institutions such as households, firms, and governments.

Uploaded by

Beatrice Martyan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 50

CLASSICAL PROGRAMMING:

THE UNCONSTRAINED CASE

D.Sc. Olena V.
Bazhenova
Classical programming

The classical programming problem is that of choosing


values of certain variables so as to maximize or minimize
a given function subject to a given set of equality
constraints.
Classical programming
The classical programming maximum problem is

or in full:
max 𝐹 ( 𝑥 ) subject ¿ 𝑔(𝑥)=𝑏
max𝑥 𝐹 (𝑥 1 , 𝑥 2 , … 𝑥𝑛 ) subject ¿
𝑥 1 , 𝑥 2 , …𝑥 𝑛

( )
¿𝑔 1 (𝑥)=𝑔1 (𝑥 1 , 𝑥 2 , ...., 𝑥 𝑛 )=𝑏 1
¿ 𝑔2 (𝑥)=𝑔 2 (𝑥1 , 𝑥 2 ,... , 𝑥𝑛 )=𝑏2
¿.
¿.
¿.
Classical programming
The n variables x1, x2, . . . , xn are the instruments,
summarized by the column vector x.

The function F(.) is the objective function, and the


m functions g1(.), g2(.), …. , gm(.) are the constraint
functions, summarized by the column vector g(.). The
constants b1 , b2 , . . . , bm are the constraint constants,
summarized by the column vector b.
Classical programming
It is assumed that the number of instruments, n,
and the number of constraints, m, are finite and that n >
m, where the difference n-m is the number of degrees of
freedom of the problem.
It is also assumed that the m +1 functions F(•),
g1(.), g2(.), …, gm(.) are given, continuously
differentiable and contain no random elements; that b
consists of given real numbers; and that x can be any
real vector, subject only to the m constraints.
The Unconstrained Case
The unconstrained case in the special case in which
there are no constraints, m = 0, and in the scalar
unconstrained case, m=0, n=1, the problem is that of
choosing the real number x so as to maximize . In such
a problem, if is a local interior maximum then, for all
neighboring points , where is an arbitrary small
variation in x:
Taylor series of a function

The Taylor series of a function is


an infinite sum of terms that are
expressed in terms of the
function's derivatives at a single
point.
Taylor series of a function
If f is a continuous function and n times differentiable in
an interval [x, x+h], then there exists some point in this
interval, denoted by x+λh for some λ∈[0,1], such that
The Unconstrained Case
Assuming is twice continuously differentiate with
continuous and finite derivatives, the function on the
right hand side of

can be expanded in a Taylor's series expansion (with


remainder term) about the point (where ) to yield
() + (+ ,
0
The Unconstrained Case
Inserting this expansion into the previous one yields the
fundamental inequality

() + (+

an inequality which must hold for any arbitrary small


variation in the instrument .
The Unconstrained Case
If is positive the fundamental inequality implies, by
dividing both sides by and taking the limit as
approaches zero:
()
The Unconstrained Case
But if is negative similar reasoning implies that:

().
The Unconstrained Case
Thus, the fundamental inequality requires as a first
order necessary condition that the first derivative
vanishes at the local maximum point:

().
The Unconstrained Case

Using the first order condition, the


fundamental inequality
()+ (+
implies, since is always positive that:
(+ .
The Unconstrained Case
Since (+ holds for all and since the second
derivative is assumed continuous, a second order
necessary condition requires that the second derivative
be negative or zero at the local maximum point:
The Unconstrained Case
Thus conditions () and are, respectively, the first
order and second order necessary conditions implied by
the existence of a local maximum at .
The Unconstrained Case
Sufficient conditions for a strict local maximum at
are the conditions that the first derivative vanish and the
second derivative be strictly negative at this point; i.e.,
the conditions
()

imply that is a strict local maximum:


The Unconstrained Case
The unconstrained vector case, m=0, n > 1, can be
treated in a similar way. The problem is

and, assuming a local maximum exists at :

,
where is an arbitrary small positive number; , is an
arbitrary variation in ; and , , …, ) is a direction in .
The Unconstrained Case
The function on the right hand side of can be
considered a function of and, expanding in a Taylor's
series expansion about the point , yields:

() +(+ (),

, is the gradient vector, and is the Hessian matrix:


The Unconstrained Case

()=((), (), …, ()),

(x)=
The Unconstrained Case
In this case we have the fundamental inequality:
() +

an inequality which must hold for all directions and all


small positive numbers .
The Unconstrained Case
Dividing both sides by and taking the limit as
approaches zero, the fundamental inequality requires as
a first order necessary condition that the gradient vector
vanishes at the local maximum point:
()

that is, a local maximum must occur at a stationary point


at which all first order partial derivatives vanish.
The Unconstrained Case
The fundamental inequality then requires as a second
order necessary condition that the Hessian matrix be
negative definite or negative semidefinite at the local
maximum point:
The Unconstrained Case
Sufficient conditions for a strict local maximum at
are the conditions that be a stationary point at which the
Hessian matrix is negative definite; i.e., the conditions:
()

imply that is a strict local maximum:


The Unconstrained Case
In the two-dimensional unconstrained problem:

for a local maximum at =, )’, the first order conditions


are that be a stationary point:
The Unconstrained Case
and the second order necessary conditions are that the
Hessian matrix be negative definite or negative
semidefinite, equivalent to the conditions on the leading
principal minors of the Hessian matrix:
The Unconstrained Case
• If the Hessian at a given point has all positive
eigenvalues, it is said to be a positive-definite matrix.
This is the multivariable equivalent of “concave up”. If
all of the eigenvalues are negative, it is said to be a
negative-definite matrix. This is like “concave down”.

• A negative semidefinite matrix is a matrix all of whose


eigenvalues are nonpositive
The Unconstrained Case
Let A be an n × n symmetric matrix.

Then:
A is positive semidefinite if and only if all the principal minors
of A are nonnegative.
A is negative semidefinite if and only if all the kth order
principal minors of A are ≤ 0 if k is odd and ≥ 0 if k is even.

Note that if a matrix is positive definite, it is certainly positive


semidefinite, and if it is negative definite, it is certainly
negative semidefinite
The Unconstrained Case
To complete the analysis of this case, a stationary point
at x* is a local minimum only if:
and holds,
and is a saddle point if does not hold.
Example 1

So, is a stationary point.


Example 1

Since we have a non-zero odd derivative, hence the point


x=1 is the inflection point.
Inflection point

Inflection points are points where the function changes


concavity, i.e. from being "concave up" to being
"concave down" or vice versa. They can be found by
considering where the second derivative changes signs.
Example 2

So, is a stationary point.


Example 2
Example 2

Thus, since we have fourth derivative (even) and it


is positive at the stationary point, then this point is a
local minimum.
Example 3

=
Example 3
Example 3
Example 3
Example 3
Example 3

Since all principal minors are non-zero, the


character of x(1) is determined by f′′(x). Since the
matrix f ′′(x(1) ) is not positive definite, then we should
consider the matrix [− f ′′(x(1) )]
Example 3
Example 3

Since the matrix [− f ′′(x(1))] is not positive


defined, then the matrix f ′′(x(1)) is not negative defined.

Hence, matrix f ′′(x(1)) is indefinite and x(1) is


saddle point.
Example 3

•
Example 3

Since all principal minors are non-zero, the


character x(2) is defined using f′′(x). Since the matrix f ′′
(x(2) ) is positive defined, then x(2) is a local minimum.
Example 4

•
=

=
Example 4
•

• M1=8>0
• M2= =8*0-(-4)(-4)=-16<0
Example 4
Application
An economy is a collection of certain institutions,
each of which faces and solves an economizing problem.
While any real economy contains a myriad of such
institutions, economic theory treats only a few idealized
but hopefully representative institutions.
Application
Among these idealized institutions are:
households: groups or persons sharing income for consumption purposes;
typically family groups.
firms: entities (proprietorships, partnerships, or corporations) producing
goods or services for sale to other firms or final consumers.
trade unions: groups of employees organized to bargain collectively with
employers for certain ends.
governments: political entities which often have important economic
functions
central banks: entities which regulate monetary process in countries.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy