UNIVERSITY OF SAN CARLOS OF GUATEMALA

UNIVERSITY CENTER OF THE EAST

BUSINESS ADMINISTRATION - SATURDAY PLAN
QUANTITATIVE METHODS III

NON-LINEAR PROGRAMMING

LUIS ERNESTO FRANCO MADRID–201342794

MARÍA FERNANDA MURCIA VIVAR–201344209
CHRISTIAN JOSUÉ PACHECO DE LA CRUZ–201342376
LUIS ROLANDO GONZÁLEZ ALVARADO–201340381
CARLOS ALBERTO CORDÓN ORTÍZ–201340368
WENS SALVADOR CHACÓN FERNÁNDEZ –200040328

CHIQUIMULA, GUATEMALA, OCTOBER 2016

 UNIVERSITY OF SAN CARLOS OF GUATEMALA
UNIVERSITY CENTER OF THE EAST
BUSINESS ADMINISTRATION - SATURDAY PLAN
QUANTITATIVE METHODS III

NON-LINEAR PROGRAMMING

RESEARCH

FOR
LUIS ERNESTO FRANCO MADRID–201342794
MARÍA FERNANDA MURCIA VIVAR–201344209
CHRISTIAN JOSUÉ PACHECO DE LA CRUZ–201342376
LUIS ROLANDO GONZÁLEZ ALVARADO–201340381
CARLOS ALBERTO CORDÓN ORTÍZ–201340368
WENS SALVADOR CHACÓN FERNÁNDEZ –200040328

Chiquimula, Guatemala, October 2016

 INDEX

INDICE .................................................................................................................... ii

INTRODUCTION

OBJECTIVES............................................................................................................ 2

CHAPTER I
NON-LINEAR PROGRAMMING (NLP)

1.1 Historia of NLP ......................................................................................... 3

1.2 Definition of PNL.......................................................................................... 4

1.3 Characteristics of PNL ................................................................................. 5

1.4 Importance from the PNL................................................................................... 6

1.5 Models of nonlinear programming............................................................... 7

1.5.1 Optimization nonlinear with constraints..................................................... 9

1.5.2 Analysis graph of nonlinear programming problems.................... 9

1.5.3 PNL with equality constraints............................................................12

1.5.4 Types of NLP problems........................................................................13

CONCLUSIONS ................................................................................................. 21

LITERATURE CONSULTED ......................................................... 22

ii
 INTRODUCTION

Most of the problems faced in the business world and real life,
they are not totally linear, they are generally characterized by being non-linear,
there lies the importance of studying this type of programming.

Non-linear programming is considered a problem-solving process of a

system of equalities and inequalities subject to a set of constraints on
a set of unknown real variables. Nonlinear programming analyzes the
problematic in which the objective or constraints are nonlinear functions.
There are several nonlinear programming problems, depending on the function of
the characteristics they possess, will be worked on as it best suits them
resolution, to achieve this the application of several algorithms is used.

A problem is a nonlinear programming problem when it has a nonlinear objective function and
one or more constraints expressed through nonlinear functions. The
Nonlinear programming is characterized by being associated with a set of
difficult problems to solve, but that can be solved through some
problem categories, among these can be mentioned: Problems
unrestrictive, restrictive problems, convex programming problems,
non-convex programming problems, separable optimization problems,
problems of geometric programming. For a better understanding of it
as exposed in the previous section, the following presents a description of
each of the topics I bought the nonlinear programming.
 OBJECTIVES

General

Knowing when a problem becomes Non-Linear Programming.

Specific

Determine which difficulties are most commonly encountered within the

Non-Linear Programming.
Establishing the importance of Non-Linear Programming.
Indicate what some of the reasons for non-linearity are in the
problems.
Determine what types of problems can arise in programming.
non-linear
 CHAPTER I
NON-LINEAR PROGRAMMING (NLP)

1.1 History of NLP

A nonlinear programming model is one where the decision variables are

they are expressed as nonlinear functions either in the objective function and/or constraints
of an optimization model. This particular feature of the models does not
linearity allows for addressing problems where there are economies or diseconomies of
scale or in general where the assumptions associated with proportionality are not
they fulfill.

The following outlines some relevant aspects and how it arises

non-linear programming

Year Event
Great mathematicians, such as Newton, Leibnitz, Bernoulli and,
17th century everything, Lagrange, who had contributed so much to the development of

and XVIII infinitesimal calculus, focused on obtaining maxima and minima

conditioned by certain functions.
Later, the French mathematician Jean Baptiste-Joseph
(1768- Fourier was the first to sense, although imprecisely, the
1830 methods of what is currently called linear programming and the
potential that derives from them.
Except for the mathematician Gaspar Monge, who in 1776
interest in problems of this kind dates back to the year
1939 to find new studies related to the methods
from the current linear programming. In that year, the Russian mathematician

1746-1818 Leonid Vitalevich Kantorovich published an extensive monograph

titled Mathematical Methods of Organization and Planning of the
production in which for the first time a is made to correspond to a
a wide range of problems a precise and well-defined mathematical theory

defined, known today as linear programming.

 4

Year Event
Joseph Fourier anticipated linear programming. Carl Friedrich Gauss
1826
solve linear equations by 'Gaussian' elimination.
Gyula Farkas conceived a method to solve systems of
1902
inequalities.
The transportation problem is formulated for the first time, studied
independently by Koopmans and by Kantorovich, reason for the
1941-1942
which is commonly known as the Koopmans problem
Kantorovftch.
George Dantzig publishes the algorithm simplex y
John von Neumann developed the theory of duality.
1947
It is known that Leonid Kantorovich also formulated the theory in the form
independent.
Another Russian mathematician, Leonid Khachiyan, designed the so-called

1979 Ellipsoid algorithm, through which he demonstrated that the problem

from linear programming
Narendra Karmarkar introduces the interior point method for
1984
solve linear programming problems.
Henry L. Gantt with production planning (diagrams of
1989
Gantt); and Ford W. Harris with models for batch calculation.
economic and inventory control.

1.2 Definition of NLP

The optimization models that describe real-life problems are

generally nonlinear and belong to the area of nonlinear optimization. In theory
in operations research, the term nonlinear programming is used for
to refer to this kind of problems.

A problem is non-linear programming when it has a non-linear objective function and

one or more restrictions expressed through non-linear functions. Typically,
a nonlinear function is a mathematical algebraic relationship where the variables
are being multiplied or are raised to an exponent different from one or
 5

where the variables are the arguments of trigonometric, logarithmic functions

or exponentials.

In response to two of the most frequent criticisms of Linear Programming

(PL), namely, the restrictiveness of the linearity hypothesis and the difficulty of defining
a single objective function, Nonlinear Programming (NLP) arises. Indeed,
An important assumption in PL is that all its functions (objective and constraints)
linear sounds. Although, in essence, this assumption holds true in the case of many
practical problems, it is often not the case. Therefore, it is necessary to address it
from Nonlinear Programming (NLP).

In general, the nonlinear programming problem consists of finding x

(x1, x2, ..., xn) for

maximize f(x),

subject to

gi(x) ≤ bi, for i = 1, 2, ..., m,

x ≥ 0,

where f(x) and gi(x) are given functions of n decision variables. Similarly,
This problem could be in their minimization case.

1.3 Characteristics of NLP

La programación no lineal se caracteriza porque está asociada a un conjunto de

difficult problems to solve. Under this name, a set of
solution techniques that are applied to various categories of problems
linear. Within those categories, the most important are the following

Unconstrained problems
Problems with restriction
Convex programming problems
Non-convex programming problems
 6

Separable optimization problems

Geometric programming problems

Among the most important difficulties faced in all cases, there can be
mention the following:

There is no universal optimization method or algorithm that solves

instead of all or the vast majority of existing problems, there is a wide
and varied collection of algorithms.
La función objetivo, las restricciones o ambas pueden ser no lineales.
The optimization methods and algorithms for NLP present
different behaviors for each type of problem.
There are cases where the optimal point is inside the feasible region.
The responsibility of selecting the best method or algorithm lies with the user.

Characteristics of the problems being analyzed:

Objective function and derivable constraints

Deterministic problems
Continuous variables
Connected solutions space
Convex objective function for the minimization problem and concave for
the problem of maximization (desirable)
Convex solution space (desirable)

1.4 Importance of NLP

It's easy to think of linear or nonlinear programming as a maze of numbers.

endless. But for many it is an integral part of mathematics that must
being studied, but in some cases it is mentioned that there is no application whatsoever in

the real world; for others, it is a way to apply optimization models to

empresas netamente productivas a nivel macro y para otras personas es un tema
which only interests systems engineers and mathematicians. However, the
the importance of linear programming lies not only in the procedure
 7

mathematical, but in the financial tool that can provide us with support
for decision-making in any organization.

The field of nonlinear programming is a key element in growing power.

of administration as a scientific method in decision-making, and despite the fact that
is based on deep mathematical foundations, supported by linear algebra
and matrices, which when conducting the studies properly can be seen as
a purely administrative tool this represents a utility that the
the administrator should take the opportunity to make decisions efficiently,
based on valid data and results for the benefit of the generating company
of goods and services.

The importance of this method stands out as mentioned earlier.

as a tool that assists the business administrator within a
dynamic administrative process. It is the administrator's duty to know and manage it,
if not in an efficient and practical way, at least to have a broad understanding of their

foundation, process and application in order to have a secure and grounded basis
to make their decisions, which are implemented for business problems
they can provide at the end, the desired objectives by the entrepreneur and to the
company the profits it wants to obtain.

1.5 Non-linear programming models

The mathematical functions or relationships that are involved in many problems

business and economic processes are not entirely linear. In fact, perhaps one can
to say that the problems of the real world fit into the strict mold of the
Linearity is the exception and not the rule.

There are many types of NLP problems, depending on the characteristics of

estas funciones, por lo que se emplean varios algoritmos para resolver los distintos
types. For certain cases where the functions have simple forms, the problems
they can be solved relatively efficiently. In some other cases,
Even solving small problems represents a real challenge.
 8

When a NLP problem has only one or two variables, it can be represented
in graphic form. If the functions are not linear, curves will be drawn instead of
lines, so the objective function and the feasible region will stop having the appearance that
they acquire in the PL. The solution does not have to be at a vertex of the region
feasible, it doesn't even have to be located on the border of this. Therefore, the
now the great simplification used in PL disappears, which allows to limit the
search for optimal solutions at the vertices. Also in NLP a
A local maximum is not necessarily a global maximum, and in general, algorithms
of NLP cannot distinguish when it is in a local optimum or in one
global. Therefore, it is crucial to know the conditions under which a
A local maximum is a global maximum in the feasible region.

In general, some important (and not necessarily distinctive) reasons for the non
linearity are: (1) non-proportional relationships decrease); (2) non-
additive (for example, when two chemical substances join, the volume
the result is not necessarily the sum of the volumes of the two substances;
y (3) the efficiencies or inefficiencies of scale (for example, when too many
workers trying to plant in the same garden start to hinder each other
and the output per worker decreases, instead of remaining constant). In
sum, any number of physical, structural, ideological relationships,
Economic and logical factors can lead to the emergence of the characteristics of not
linearity in a model.

It is worth repeating that, although nonlinear phenomena are common, the

optimization of nonlinear models is much more difficult than that of models
linear. For example, unlike PL, it cannot be assumed that the
nonlinear optimization procedure of Solver will always manage to find the
optimal solution for all nonlinear models. Adding to this difficulty the
the fact that, in many contexts, linear models provide good
approaches to nonlinear models, for this reason it can be understood why
linear models are as popular as PL.
 9

1.5.1 Nonlinear optimization with constraints

In a management-oriented decision-making environment, it is advisable

optimize an objective function subject to constraints. These constraints take on the
form of equalities and/or inequalities in mathematics, as in the case of the
linear programming, except that now no condition of linearity is assumed.
Thus, the general model of mathematical programming can be expressed in terms of
symbolic as indicated below.

1.5.2 Graphical analysis of nonlinear programming problems

Just like with the PL, two-dimensional geometry can be used to capture.
better this problem. For example, graphical analysis is carried out to solve this
specific problem:

Note that everything in this model is linear, except for the first constraint. It is said
that a model is not linear when at least one of the constraint functions
either the objective function or both are non-linear. Therefore, it is appropriate to say that the

The previous model is a nonlinear programming (NLP). In the following figure,

You can observe this problem graphically.
 10

La región factible.A fin de usar el método gráfico para resolver este problema, se
it will proceed in the same way as in linear programming. First trace the
set of points that simultaneously satisfy all the constraints. Equals
that in the case of the PL, this series of points is called the restricted set or the
feasible region. This set represents the permissible decisions. To find
a permitted decision that maximizes the objective function, the search is for the "most
"ascending" (that is, the one with the highest value) that still touches some point of the set

restricted. That point of contact will be an optimal solution (which frequently is

just call it a solution) to the problem.

In the previous figure, it can be seen that the non-linear constraint exhibits curvature in
the limit of the restricted set. The feasible set is no longer a polyhedron (i.e.,
a figure of flat faces, defined by linear inequalities) as in the case of
the LP, and the optimal solution is not found at a vertex. It is necessary to remember that
In the case of the PL, the graphical analysis allowed the identification of the active constraints.

at an optimal vertex, and the exact solution was obtained after solving the system
of two equations with two unknowns. In general, this method does not work for the
nonlinear case. As seen in the presented figure, there is only one restriction
active.

Optimums that are not at the vertex. Another example of NLP appears in the figure that
The following presents a hypothetical model of maximization.
 11

restricted with non-linear inequality. In this figure, all the constraints are
linear and, therefore, its set is a polyhedron. However, the objective function is
non-linear and, once again, it is evident that the solution does not appear at a vertex. Of

Indeed, the solution for some nonlinear objective functions may not even be
siquiera en el límite de la región factible. Por supuesto, podría aparecer una solución
at a vertex, but the important thing is that this does not represent a property
guaranteed, as in the linear model.

This fact has important algorithmic implications. It means that, in the case
non-linear, a "vertex search" method, like the method cannot be used
simplex used by Solver to find a solution in LP models. With
this restriction complicates the resolution process enormously.

Example 2. Given the present mathematical model, the feasible region is located,
agreement with the respective restrictions. It is worth mentioning that its solution is the

interior point, (3, 3), of the feasible region.

 12

If a linear programming problem has no constraints, the fact that the objective function
A concave sea guarantees that a local maximum is a global maximum. (In the same way,
a convex objective function ensures that a local minimum is a global minimum.) If
there are restrictions, an additional condition is needed to provide this guarantee, namely,
that the feasible region is a convex set. For this reason, the sets
Convex functions play a fundamental role in nonlinear programming.

A convex set is simply a set of points such that, for each

For pairs of points in the collection, the line segment that connects them is completely
content in it. Thus, the feasible region in the previously presented problem,
it is a convex set. In fact, the feasible region of any other problem
The PL is a convex set. In general, the feasible region of an NLP problem
it is a convex set as long as all the functions gi(x) are convex.
to ensure that a local maximum is a global maximum in a NLP problem with
restrictions g1(x)≤b1, g2(x)≤b2, ..., gm(x)≤bm and x≥0, the objective function

f(x) must be concave and each gi(x) must be convex. A problem of this type is
It is called a convex programming problem and is one of the most important classes.
of NLP.

1.5.3 NLP with equality constraints

Many nonlinear problems in management and economics have the following

form:

That is to say, the goal is to maximize or minimize an objective function.

variables, subject to a set of m (m<n) equality constraints.
The following are three examples.

Example 1. A manufacturer has the possibility of producing a product in any

of the two machines. Suppose that 1it is the amount produced in machine 1 and

2the quantity produced in machine 2. Let it be

 13

2
1 1+ 1 1 production cost on machine 1

2
2 2+ 2 2production cost on machine 2

Find the values of 1y 2with which the total cost is minimized,

respecting the requirement that total production reaches a specific value,
for example R. The formulation of this problem is

Min 1 1+ 2 2
1 1+ 2 2+ 2 2

s.a. 1+ 2=R

Example 2. Suppose that 1 , 2y 3they are the prices of three goods and be

available budget (i.e., B is a specific constant). Let 1 , 2y 3

1 2 3
previously determined constants and is represented with 1+ 2+ 3the

derived utility of consumption 1units of good 1, 2units of good 2 and 3

units of the good 3.

Find the consumption mix that maximizes utility while respecting the constraint.
budgetary. The formulation of this problem is

3
Max 1 1 + 2 +2 3

s.a. 1 1+ 2 2+ 3 3=R

1.5.4 Types of NLP problems

Non-linear programming problems appear in many different forms.

Unlike the simplex method for linear programming, there is no
algorithm that solves all these special types of problems. Instead, it
they have developed algorithms for some classes (special types) of problems of
nonlinear programming.

A continuación se introducirán brevemente las clases más importantes de

problems that occur in nonlinear programming.
 14

a) Unconstrained optimization

Unconstrained optimization problems have no constraints, so the

the objective function is simply,

Maximize ( )

about all the values of x = ( 1 2… , The necessary condition for a

specific solution = ∗ it is optimal when ( ) it is a differentiable function

When ( ) it is concave, this condition is also sufficient, so that the

the obtaining of x* reduces to solving the system of the n equations that
they obtained by setting the n partial derivatives equal to zero.
Unfortunately, when it comes to nonlinear functions ( ), you are
equations are also often nonlinear, in which case it is unlikely that they will
can obtain a simultaneous analytical solution.

When a variable , it has a non-negativity restriction, , >=0, the condition

necessary and—perhaps—sufficient previous changes slightly to

for each of this type. This condition is illustrated in the following figure, where the
the optimal solution of a problem with a single variable is x = 0 even when the
the derivative there is negative and not zero. As this example has a concave function
to maximize subject to a non-negativity constraint, that its derivative is
less than or equal to 0 at x = 0 is a necessary and sufficient condition for x = 0
optimal sea.
 15

A problem that has some non-negativity constraints and that does not have
functional restrictions are a special case (m = 0) of the following class of
problems.

b) Linearly constrained optimization

Linearly constrained optimization problems are characterized by

restrictions that fully comply with linear programming, so that
all constraint functions ( ) linear ones, but the objective function ( )
it is nonlinear. The problem simplifies significantly if only has to take
consider a non-linear function along with a feasible region of linear programming.
Several special algorithms have been developed based on an extension of
método símplex para analizar la función objetivo no lineal. Un caso especial
The important topic described below is quadratic programming.

c) Quadratic programming

Once again, quadratic programming problems have linear constraints,

but now the objective function ( ) it must be quadratic. So, the only
The difference between these and a linear programming problem is that some
terms of the objective function include the square of a variable or the product
 16

of two variables. Many algorithms have been developed to handle this case,
with the additional assumption that ( ) is concave. The quadratic programming is
very important, partly because formulations of this type arise in a way
natural in many applications. For example, the selection of a portfolio with
risky investments, fits this format. However, another reason why
It is important that when solving general constrained optimization problems
linearly one can obtain the solution of a succession of approximations of
quadratic programming.

d) Convex programming

Convex programming encompasses a wide class of problems, among which,

As special cases, all the previous types can be mentioned when ( )
It is a concave function that needs to be maximized. The assumptions are:

( ) it is concave.
Each of the ( ) it is convex.

These assumptions are sufficient to ensure that a local maximum is a maximum.

global. (If, on the contrary, the goal were to minimize ( )subject to either ( )≤
to a– ( )≥ for 1, 2, ... , the first assumption would change to that ( ) should
be a convex function, since that is what is required to ensure that a
a local minimum is a global minimum.

e) Separable programming

Separable programming is a special case of convex programming, in

where the additional assumption is:

All functions ( ) y ( ) they are separable.

A separable function is a function in which each term includes a single

variable, so the function can be separated into a sum of functions of
individual variables. For example, if ( ) it is a separable function, it can be
express as
 17

( ) =∑ ( )
=1

where each ( ) include only the terms with In the terminology of

linear programming, separable programming problems satisfy the
additivity assumptions, but violate the proportionality assumption when
any of the functions ( ) they are nonlinear functions.

To illustrate, the objective function considered in the previous figure

2 2
( 1, )2= 126x1-9x1+ 182 2− 13 2

it is a separable function because it can be expressed as

( 1, )2= 1 ( 1+
() ) 22

2
where 1 ( 1=
) 126x1- 9x 1 2= 182
2( )2− 13 are each22 functions of
a single variable 1y 2 , respectively. It is important to distinguish these
problems of others in convex programming, since any problem of
separable programming can be closely approximated by one of
linear programming and, therefore, the efficient simplex method can be applied.
 18

f) Non-convex programming

Non-convex programming includes all non-linear programming problems.

that do not meet the assumptions of convex programming. In this case, even
when successful in finding a local maximum, there is no guarantee that it is
also a global maximum. Therefore, there is no algorithm that guarantees
to find an optimal solution for all these problems; however, there are
some algorithms quite suitable for finding local maxima, in
especially when the shapes of the nonlinear functions do not deviate too much
of those that were supposed for convex programming.

Certain specific types of non-convex programming problems can be

solve without much difficulty using special methods.

g) Geometric programming

When nonlinear programming is applied to engineering design problems,

many times the objective function and the constraint functions take the form

( )=∑ ( ),
=1

Where

( ) = 1
1
2
2...
, para = 1,2, . . . , .

In such cases, y they often represent physical constants, while

that the they are the design variables. These functions are generally neither
neither concave nor convex, so convex programming techniques do not
they can be directly applied to these problems of geometric programming.
However, there is an important case in which the problem can be transformed.
in an equivalent convex programming problem. This case is one in which
that all the coefficients the outputs of each function are strictly positive, that is,
the functions are generalized positive polynomials (now called posinomials),
and the objective function must be minimized. The equivalent problem of
 19

convex programming with decision variables 1 , 2, . . ., is obtained when

establish

= for = 1,2, . . . , .

in the entire original model, so that an algorithm can now be applied

convex programming. Another solution procedure has been developed for
resolve these subpolyhedral programming problems, just like for problems
of geometric programming of other types.

h) Fractional programming

Suppose that the objective function is in the form of a fraction, that is,
the ratio or quotient of two functions,

1 ( )
Maximize ( )=
2 ( )

These fractional programming problems arise, for example, when

maximize the production rate per man-hours employed
(productivity), or the gain between the invested capital (rate of return), or the
expected value divided by the standard deviation of some measure of
performance of an investment portfolio (return/risk). Formulations have been made
some special solution procedures for certain forms of

1( ) 2( ) .

When possible, the most direct approach to solving a problem of

fractional programming is transforming it into an equivalent problem of some type
standard that has an efficient procedure. To illustrate this approach,
suppose that ( ) it is in the form of fractional linear programming

+ 0
( )= ,
+ 0
 20

where c and d are row vectors, x is a column vector and 0y 0they are scalars.

Also suppose that the constraint functions ( ) linear, that is, the
Matrix constraints are Ax ≤ b and x ≥ 0.

Under some additional weak assumptions, the problem can be transformed into
an equivalent linear programming problem if it is established

1
= y = ,
+ 0 + 0

so that = / . This result leads to

Maximize = + 0 ,

Subject to

− ≤ 0,
+ 0 = 1,

≥ 0, ≥ 0,

what can be solved with the simplex method. In general terms, it can be
use the same type of transformation to convert a programming problem
fractional with 1( )concave 2 (convex
) and ( ) convex, in a problem
equivalent of convex programming.
 CONCLUSIONS

A problem becomes Non-Linear Programming when it has a function

nonlinear objective and one or more constraints expressed through non-functions
literals.
Some of the difficulties that are most encountered in programming do not
Linear means that there is no universal optimization method or algorithm that
solve most problems, the methods and algorithms present
different behaviors for each type of problem, the variables are
continues, the responsibility of selecting the best method or algorithm falls
in the user, etc.
The importance of nonlinear programming lies not only in the procedure.
mathematician, but in the financial tool it can provide us as
support for decision-making in any organization.
Among the reasons why a problem is nonlinear are that the relationships
non-proportional decreases, non-additive relationships and efficiencies or
scale inefficiencies.
Linear programming problems can take many forms.
different, with the most common being unconstrained and unconstrained optimization,
as well as convex and non-convex programming, quadratic programming,
separable programming, geometric programming, and fractional programming.
 LITERATURE CONSULTED

Avriel, Mordecai. Nonlinear. Programming: Analysis and Methods. Dover

Publishing, 2003.

Bazaraa, Mokhtar S. and Shetty, C. M. Nonlinear programming. Theory and

algorithms. John Wiley & Sons, 1979.

Bertsekas, Dimitri P. Nonlinear Programming: 2nd Edition. Athena Scientific, 1999.

Calix Ng, Carlos Amadeo. Linear programming, as a basic tool in the

Planning and Decision Making. Bachelor's thesis in Administration
empresas, Escuela de Administración de Empresas : Universidad Mariano
Gálvez, 1982.

Díaz, Gloria Marlene. Linear programming as a tool for decision-making

financial decisions. Santo Tomas University, 2002.

Eppen, G.D.; Gould, F.J.; Schmidt, C.P.; Moore, Jeffrey H. y Weatherford, Larry R.
Operations research in administrative science. Mexico: Prentice-
Hall, 5th edition, 2000.

Escobar Zuluaga, Antonio Hernando; Gallego Rendón, Ramón Alfonso and Romero
Lázaro, Rubén Augusto. Nonlinear Programming. Technological University of
Pereira, Colombia, 2014.

Hillier, Frederick S. and Lieberman, Gerald J. Introduction to Research

operations. Mexico: McGraw-Hill Publishing, Ninth edition. 2010.

Merino Maestre, María. Classical Optimization Techniques Part I: Programming

Linear and Non-Linear. Faculty of Science and Technology, Department of
Applied Mathematics and Statistics and Operations Research. Available at:
[Link]

Nocedal, Jorge and Wright, Stephen J. Numerical Optimization. Springer, 1999.

Taha, Hamdy A. Operations research. Mexico: Pearson Education.

Ninth edition, 2012.