BARON user manual v. 15.6.

5
June 5, 2015

Nick Sahinidis, The Optimization Firm, LLC, niksah@minlp.com, http://www.minlp.com

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1 Licensing and software requirements . . . . . . . . . . . . . . . . . . . . 2
2 Model requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Allowable nonlinear functions . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Variable and expression bounds . . . . . . . . . . . . . . . . . . . . . . . 3
3 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4 BARON input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.1 Usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.2 Input grammar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.3 The options section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.4 The problem data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.5 Error messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.6 Sample input file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.7 Other ways to access BARON . . . . . . . . . . . . . . . . . . . . . . . . 9
5 BARON output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.1 BARON screen output . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.2 Termination messages, model and solver statuses . . . . . . . . . . . . . 11
5.3 BARON solution output . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
6 Some BARON features . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6.1 No starting point is required . . . . . . . . . . . . . . . . . . . . . . . . 15
6.2 Finding a few of the best or all feasible solutions . . . . . . . . . . . . . 15
6.3 Using BARON as a multi-start heuristic solver . . . . . . . . . . . . . . 17
6.4 Systematic treatment of unbounded problems . . . . . . . . . . . . . . . 18
6.5 Systematic treatment of infeasible problems . . . . . . . . . . . . . . . . 18
6.6 Handling of complementarity constraints . . . . . . . . . . . . . . . . . . 19
6.7 Parallel capabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2 BARON user manual v. 15.6.5

7 The BARON options . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

7.1 Termination options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
7.2 Relaxation options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
7.3 Range reduction options . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
7.4 Tree management options . . . . . . . . . . . . . . . . . . . . . . . . . . 23
7.5 Local search options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
7.6 Output options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
7.7 Subsolver options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
7.8 Licensing options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
7.9 Other options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
8 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

1 Introduction
The Branch-And-Reduce Optimization Navigator (BARON) is a computational system for the
global solution of algebraic nonlinear programs (NLPs) and mixed-integer nonlinear programs
(MINLPs).
While traditional NLP and MINLP algorithms are guaranteed to provide global optima only
under certain convexity assumptions, BARON implements deterministic global optimization al-
gorithms of the branch-and-bound type that are guaranteed to provide global optima under fairly
general assumptions. These assumptions include the existence of finite lower and upper bounds
on nonlinear expressions in the NLP or MINLP to be solved.
BARON implements algorithms of the branch-and-bound type, enhanced with a variety of con-
straint propagation and duality techniques for reducing ranges of variables in the course of the
algorithm.
Parts of the BARON software were created at the University of Illinois at Urbana-Champaign
and Carnegie Mellon University.

1.1 Licensing and software requirements

The demo version of BARON is freely available from The Optimization Firm and can be down-
loaded from http://www.minlp.com/download. This code can handle problems with up to 10
variables, 10 constraints, and 50 nonlinear operations. In order to use BARON for larger prob-
lems, users will need to have a valid BARON license. The Optimization Firm provides licenses
that permit users to use BARON directly on any Windows or Linux platform as well as under
MATLAB and YALMIP. In addition, BARON distributors AIMMS, AMPL and GAMS provide
licenses for using BARON under their modeling systems.
BARON user manual v. 15.6.5 3

The software includes the Coin solvers CLP/CBC, FilterSD, and IPOPT for solving BARON’s
linear/integer programming (LP/MIP) and nonlinear programming (NLP) subproblems. In ad-
dition, BARON can utilize IBM’s ILOG CPLEX for solving LP/MIP subproblems if CPLEX is
installed on the user’s computer as a dynamic library. Additional licensed LP/MIP and NLP
solvers are available under the AIMMS, AMPL and GAMS versions of BARON and may expe-
dite convergence. Current valid LP/MIP subsolvers for BARON are CLP/CBC, CPLEX, and
XPRESS. Current valid NLP subsolvers are CONOPT, FilterSD, IPOPT, MINOS, and SNOPT.

2 Model requirements
BARON addresses the problem of finding global solutions to general nonlinear and mixed-integer
nonlinear programs:

min f (x)
s.t. g(x) ≤ 0
x∈X

where f : X → R, g : X → Rm , and X ⊂ Rn . The set X may include integer restrictions and the
constraints may include complementarity relationships. The types of functions f and g currently
handled by BARON are discussed below.

2.1 Allowable nonlinear functions

In addition to multiplication and division, BARON can handle nonlinear functions that in-
volve exp(x), ln(x), xα for real α, and β x for real β. AIMMS/BARON, AMPL/BARON and
GAMS/BARON automatically handle |x| and xy , where x and y are variables; otherwise, suitable
transformations discussed below can be used. Currently, there is no support for other functions,
including the trigonometric functions sin(x), cos(x), etc.

2.2 Variable and expression bounds

Nonlinear expressions in the mathematical program to be solved must be bounded below and/or
above. It is important that finite lower and upper bounds be provided by the user on as many
problem variables as possible. However, providing finite bounds on variables alone may not
suffice to guarantee finite bounds on nonlinear expressions arising in the model. For example,
consider the term 1/x for x ∈ [0, 1], which has finite variable bounds, but is unbounded. It is
important to provide bounds for problem variables that guarantee that the problem functions
are finitely-valued in the domain of interest. If the user model does not include variable bounds
that guarantee that all nonlinear expressions are finitely-valued, BARON will attempt to infer
appropriate bounds from problem constraints. If this step fails, global optimality of the solutions
provided may not be guaranteed.
4 BARON user manual v. 15.6.5

3 Installation
If you intend to use BARON under GAMS, AIMMS, AMPL, or YALMIP, the BARON software
already comes installed with your modeling system. If you intend to access the stand-alone
version of BARON, entirely on its own or via MATLAB, place the BARON executable and
license in your PATH. Simply placing them under your MATLAB PATH may not suffice. In
order to use BARON under MATLAB, follow the additional installation instructions that come
with the MATLAB/BARON interface.

4 BARON input
There are various ways for passing an optimization problem to BARON:

• Directly, using the BARON modeling language.

• Indirectly, using one of the available BARON interfaces under AIMMS, AMPL, GAMS,
MATLAB, or YALMIP.

In this section, we describe the BARON modeling language in detail.

4.1 Usage
BARON provides a high-level modeling language capable of reading a mixed-integer nonlinear
optimization model in a relatively simple format. Input is provided in the form of a text file. Even
though not required, it is strongly recommended that all BARON input files have the extension
.bar. Let the input file be called test.bar and let the name of the BARON executable be
baron. Then, issuing the command

baron test

or

baron test.bar

results in BARON parsing the file and solving the problem.

4.2 Input grammar

The following rules should be adhered to while preparing a BARON input file:

• All statements should be terminated by a semicolon (;).

• Reserved words must appear in uppercase letters.

BARON user manual v. 15.6.5 5

• Variable names can be in lower or upper case. The parser is case sensitive, i.e., X1 and x1
are two different variables.
• Variable names should be no longer than 15 characters.
• Variable names must start with a letter.
• With the exception of underscores ( ), non-alphanumeric characters such as hyphens (-)
are not permitted in variable names.
• Any text between // and the end of a line is ignored (i.e., it is treated as a comment).
• The signs “+”, “-”, “*” and “/” have their usual meaning of arithmetic operations.
• “ˆ” is the power/exponentiation operator where, if the base is a negative constant, the
exponent must be an integer. Note that the order of operations may be different in different
computing environments:
– xˆyˆz = (xˆy)ˆz in GAMS, MATLAB, and Excel
– xˆyˆz = xˆ(yˆz) in Fortran, AMPL, BARON, Mathematica.
• The exponential function is denoted as exp().
• The natural logarithm is available as log() as well as ln(). To enter log10 () in the model,
use the transformation log10 (x) = log10 (e) ∗ log(x) = 0.4342944819032518 ∗ log(x).
• BARON does not allow xy , where x and y are both variables. It is permissible to have
either x or y as a variable in this case but not both. The following reformulation can be
used around this: xy = exp(y ∗ log(x)). This reformulation is done automatically when
BARON is used under AIMMS, AMPL, GAMS or MATLAB.
• BARON does not allow the use of absolute values |x| in the model file. However, this func-
tion can be modeled equivalently as |x| = (x2 )0.5 . This reformulation is done automatically
when BARON is used under AIMMS, AMPL, GAMS, or MATLAB.
• Parentheses (“(” and “)”) can be used in any meaningful combination with operations in
mathematical expressions.

The input file is divided into two sections: the options and the problem data sections.

4.3 The options section

This section is optional. If used, it should be placed before any other programmatic statements.
Any of BARON’s algorithmic options can be specified here. This section has the following form:

OPTIONS {
<optname1>: <optvalue1>;
<optname2>: <optvalue2>;
<optname3>: <optvalue3>;
}
6 BARON user manual v. 15.6.5

The names and corresponding values of the BARON options are described in detail in Section 7.
Options not specified here take their default values. Instead of OPTIONS, the word OPTION can
also be used.

4.4 The problem data

This section contains the data relating to the particular problem to be solved. The section can
be divided into the following parts. Note that the words EQUATIONS, ROWS, and CONSTRAINTS are
used interchangeably.

• Variable declaration: All variables used in the problem have to be declared before they
are used in equations. Variables can be declared as binary, integer, positive or free using the
keywords BINARY VARIABLES, INTEGER VARIABLES, POSITIVE VARIABLES, and VARIABLES
respectively. In these keywords, VARIABLE or VAR may be used instead of VARIABLES and
the underscore may be replaced by a space. All discrete (binary and integer) variables
should be declared before any continuous variables. A sample declaration is as follows:

BINARY_VARIABLES y1, y2; // 0-1 variables

INTEGER_VARIABLES x3, x7; // discrete variables
POSITIVE_VARIABLES x1, x4, x6; // nonnegative variables
VARIABLE x5; // this is a free variable

• Variable bounds (optional): Lower and upper bounds on previously declared variables
can be declared using the keywords LOWER BOUNDS and UPPER BOUNDS, respectively. The
word BOUND can be used instead of BOUNDS. A sample bounds declaration follows:

LOWER_BOUNDS{
x7: 10;
x5: -300;
}

UPPER_BOUND{
x4: 100;
}

• Branching priorities (optional): Branching priorities can be provided using the keyword
BRANCHING PRIORITIES. The default values of these parameters are set to 1. Variable vio-
lations are multiplied by the user-provided priorities before a branching variable is selected.
A sample branching priorities section follows:

BRANCHING_PRIORITIES{
x3: 10;
x5: 0; }

The effect of this input is that variable x3 will be given higher priority than all others,
while variable x5 will never be branched upon.
BARON user manual v. 15.6.5 7

• Equation declaration: An identifier (name) corresponding to each equation (constraint)

has to be declared first. The keywords EQUATION and EQUATIONS can be used for this
purpose. A sample equation declaration is shown below.

EQUATIONS e1, e2, e3;

The naming rules for equations are the same as those for variables, i.e., all equation names
are case-sensitive and should begin with a letter.

• RELAXATION_ONLY_EQUATIONS <list equation names>;

This equation declaration can be used to specify constraints to be used for relaxation
construction only. This is optional and must follow after the EQUATIONS declaration and
before the equation definitions.

• CONVEX_EQUATIONS <list equation names>;

This equation declaration can be used to specify constraints that are convex. This is op-
tional and must follow after the EQUATIONS declaration and before the equation definitions.

• Equation definition: Each equation (or inequality) declared above is written in this
section of the input file. The equation is preceded by its corresponding identifier. The
bounds on the equations can be specified using the symbols == (equal to), <= (less than
or equal to) and >= (greater than or equal to). Both <= and >= can be used in the same
equation. A sample equation definition is shown below.

e1: 5*x3 + y2 - 3*x5^3 >= 1;

e2: y1 + 2*x4 - 2*x7 == 25.7;
e3: -20 <= x4 + 2*y1*x3 + x6 <= 50;

Any variables must appear only on one side of the relational operator. That is, the “left-
hand side” and the “right-hand side” should be pure numbers or expressions involving
constants but no variables.

• Objective function: BARON optimizes a given objective function. This can be declared
using the OBJ and the minimize or maximize keywords. A sample objective definition is
shown below:

OBJ: minimize 7*x3 + 2*x6;

• Starting point (optional): A starting point can be optionally specified using the keyword
STARTING POINT as follows:

STARTING_POINT{
x1: 50;
x4: 100;
x7: 300;
}
8 BARON user manual v. 15.6.5

4.5 Error messages

Any errors in the input file are reported in the form of “warnings” and “errors.” BARON tries to
continue execution despite warnings. In case the warnings and/or errors are severe, the program
execution is stopped and the line where the fatal error occurred is displayed. The input file
should be checked even if the warnings are not severe, as the problem might have been parsed
in a way other than it was intended to be.

4.6 Sample input file

A sample input file for BARON is shown below:

// This is a gear train design problem taken from the GAMS test library
//
// A compound gear train is to be designed to achieve a specific
// gear ratio between the driver and driven shafts. The objective
// of the gear train design is to find the number of teeth of the
// four gears and to obtain a required gear ratio of 1/6.931.
//
// The problem originated from:
// Deb, K, and Goyal, M, Optimizing Engineering Designs Using a
// Combined Genetic Search. In Back, T, Ed, Proceedings of the
// Seventh International Conference on Genetic Algorithms. 1997,
// pp. 521-528.

INTEGER_VARIABLES i1,i2,i3,i4; // number of teeth in each of the gears

LOWER_BOUNDS{
i1: 12;
i2: 12;
i3: 12;
i4: 12;
}

UPPER_BOUNDS{
i1: 60;
i2: 60;
i3: 60;
i4: 60;
}

EQUATIONS e2,e3; // symmetry constraints

e2: - i3 + i4 >= 0;
e3: i1 - i2 >= 0;
BARON user manual v. 15.6.5 9

// the objective aimms to make the reciprocal of the

// gear ratio as close to 6.931 as possible.
// an ideal design will have an objective equal to 1.

OBJ: minimize (6.931 - i1*i2/(i3*i4))^2 + 1;

STARTING_POINT{
i1: 24;
i2: 24;
i3: 24;
i4: 24;
}

Additional examples can be found at http://www.minlp.com/download.

4.7 Other ways to access BARON

For information on how to access BARON under MATLAB, see the BARON/MATLAB interface
manual at http://www.minlp.com/downloads/matbar/matbar.zip. For information on how to
access BARON under AIMMS, AMPL, GAMS or YALMIP, consult the corresponding web sites
of these modeling systems.

5 BARON output

5.1 BARON screen output

The screen output below is obtained for the MINLP model gear.bar.

===========================================================================
BARON version 15.6.5. Built: LNX-64 Fri Jun 5 09:12:41 EDT 2015

If you use this software, please cite:

Tawarmalani, M. and N. V. Sahinidis, A polyhedral
branch-and-cut approach to global optimization,
Mathematical Programming, 103(2), 225-249, 2005.

BARON is a product of The Optimization Firm, LLC. http://www.minlp.com/

Parts of the BARON software were created at the
University of Illinois at Urbana-Champaign.
===========================================================================
No BARON license file found in user PATH. Continuing in demo mode.
Model size is allowable within BARON demo size.
This BARON run may utilize the following subsolver(s)
For LP/MIP: ILOG CPLEX
10 BARON user manual v. 15.6.5

For NLP: COIN IPOPT with MUMPS and METIS, FILTERSD

===========================================================================
Starting solution is feasible with a value of 36.1767610000
Doing local search
Solving bounding LP
Starting multi-start local search
Done with local search
===========================================================================
Iteration Open nodes Time (s) Lower bound Upper bound
1 1 0.02 1.00000 36.1768
* 1+ 1 0.02 1.00000 35.1952
* 1+ 1 0.02 1.00000 4.72876
* 1+ 1 0.02 1.00000 3.29321
1 1 0.02 1.00000 3.29321
* 2 2 0.02 1.00000 1.06987
* 3+ 1 0.02 1.00000 1.02321
* 6 3 0.03 1.00000 1.01273
* 17 0 0.04 1.00000 1.00006
17 0 0.04 1.00000 1.00006

Cleaning up

*** Normal completion ***

Wall clock time: 0.04

Total CPU time used: 0.04

Total no. of BaR iterations: 17

Best solution found at node: 17
Max. no. of nodes in memory: 11

All done
===========================================================================

The solver first tests feasibility of the user-supplied starting point. This point is found to be feasi-
ble with an objective function value of 36.176761. BARON subsequently performs a randomized
local search procedure, which, in this case, does not improve the incumbent. Execution then
proceeds with branch-and-bound, with information printed every 1,000,000 branch-and-bound
iterations and every 30 seconds. Additionally, information is printed at the end of the root node,
whenever the value of the incumbent is improved by at least 10−5 , and at the end of the search.
During branch-and-bound search, a star (*) in the first position of a line indicates that a better
feasible solution was found that improves the value of previous best known solution by at least
10−5 . The log fields include the iteration number, number of open branch-and-bound nodes, the
time taken thus far, the lower bound, and the upper bound for the problem. The log output
fields are summarized below:
BARON user manual v. 15.6.5 11

Field Description
Iteration The number of the current iteration. A plus (+) following the iteration num-
ber denotes reporting while solving a probing (as opposed to a relaxation)
subproblem of the corresponding node.
Open Nodes Number of open nodes in branch-and-reduce tree.
Time Current computational time in seconds. CPU time is reported for single-
threaded jobs and wall clock time is reported for multi-threaded jobs.
Lower Bound Current lower bound on the model.
Upper Bound Current upper bound on the model.

Once the branch-and-reduce tree is searched, the best solution is isolated and a corresponding
dual solution is calculated. Then, the total number of branch-and-reduce iterations (number of
search tree nodes) is reported, followed by the node where the best solution was identified (a -1
indicates preprocessing as explained in the next section on termination messages).

5.2 Termination messages, model and solver statuses

Upon normal termination, BARON will report the node where the optimal solution was found.
We refer to this node as nodeopt. The log message is of the form:

Best solution found at node: (nodeopt)

where nodeopt can take the following values:



 −3 no feasible solution found,
−2 the best solution found was the user-supplied,

nodeopt =

 −1 the best solution was found during preprocessing,
i the best solution was found in the ith node of the tree.


In addition to reporting nodeopt, upon termination, BARON will issue one of the following
statements:
• *** Normal completion ***. This is the most desirable termination status. The problem
has been solved within tolerances in this case. If BARON returns a code of -3, then no
feasible solution exists.
• *** Heuristic termination ***. While global optimality is not guaranteed in this case,
BARON will terminate with this message when (a) a feasible solution has been found and
(b) the progress of lower/upper bounds satisfies the heuristic termination criterion set by
the user through BARON’s DeltaTerm option.
• *** User did not provide appropriate variable bounds ***. The user will need to
read the BARON summary file for pointers to variables and expressions with missing bounds.
The model should be modified in order to provide bounds for variables and intermediate
expressions that make it possible for BARON to construct reliable relaxations. Even though
relaxation bounds are printed on the screen to give the user a feeling for convergence, these
bounds may not be valid for the problem at hand. This message is followed by one of the
following two messages:
12 BARON user manual v. 15.6.5

• *** Infeasibility is therefore not guaranteed ***. This indicates that, be-
cause of missing bounds, no feasible solution was found but model infeasibility was
not proven.
• *** Globality is therefore not guaranteed ***. This indicates that, because
of missing bounds, a feasible solution was found but global optimality was not proven.

• *** Max. allowable nodes in memory reached ***. The user will need to increase
the physical memory of the computer or change algorithmic options, such as branching and
node selection rules, to reduce the size of the search tree and memory required for storage.
• *** Max. allowable BaR iterations reached ***. The user will need to increase the
maximum number of allowable iterations. The BARON option is MaxIter.
• *** Max. allowable time exceeded ***. The user will need to increase the maximum
of allowable time. The BARON option is MaxTime.
• *** Problem is numerically sensitive ***. BARON is designed to automatically
handle problems with numerical difficulties. However, for certain problems, the global
optimum is numerically sensitive. This occurs, for instance, when the objective function
value varies significantly over small neighborhoods of points that are strictly outside the
feasible region but are nonetheless feasible within numerical tolerances. When BARON
returns this message, the “Best possible” reported on the objective is likely correct.
• *** Search interrupted by user ***. The run was interrupted by the user (Ctrl-C).
• *** Insufficient Memory for Data structures ***. More memory is needed to set
up the problem data structures. The user will need to increase the physical memory
available on the computer in order to accommodate problems of this size.
• *** A potentially catastrophic access violation just took place. In the unlikely
event of an access violation, BARON will terminate the search and return the best known
solution. Please report problems that lead to this termination condition to Nick Sahinidis
(niksah@minlp.com).

5.3 BARON solution output

When BARON is used under AIMMS, AMPL, GAMS, MATLAB or YALMIP, the corresponding
BARON interface brings BARON results into these modeling environments; these users can skip
this section. For those users who choose to use BARON outside of these modeling systems,
BARON’s solution must be read from three output files:

• The results file provides the results. Each solution found by BARON is reported in this
file as soon as it is found. Variable values and dual values for variables and constraints
are printed in the order in which variables and constraints are defined in the BARON file.
At the end of this file, a termination message, such as “*** Normal Completion ***” is
printed, followed by the best solution point in two different formats, the last of which makes
use of the variable names used in the BARON file.

• The summary file contains the information that goes to the screen. In addition, it provides
information on missing bounds, if any.
BARON user manual v. 15.6.5 13

• The time file contains a single line with concise information on the solution, including a
breakdown of iterations and times (the same information is available at the bottom of the
summary file as well.)

As detailed in Section 7, the user has full control on whether any of these files will be written
or not. In addition, the user can specify the names and/or paths of these output files. The time
file should be read first after BARON’s termination in order to obtain information regarding
termination status. This file contains a single line with the following information:

• ProName.

• The number of constraints of the optimization problem.

• The number of variables of the optimization problem.

• The number of constraints in one of BARON’s core reformulations of the optimization

problem.

• The number of variables in one of BARON’s core reformulations of the optimization prob-
lem.

• BARON’s lower bound for the global optimum of the problem.

• BARON’s upper bound for the global optimum of the problem.

• BARON’s solver status, which can take one of the following values:

1. If normal completion occurred, i.e., the problem was solved within tolerances.
2. If there is insufficient memory to store the number of nodes required for this search
tree (increase physical memory or change algorithmic options).
3. If the maximum allowed number of iterations was exceeded (increase maxiter).
4. If the maximum allowed time was exceeded (increase maxtime).
5. If the problem is numerically sensitive.
6. If the run was interrupted by user (Ctrl-C)
7. If there was insufficient memory to setup BARON’s data structures (increase physical
memory).
8. This return code is reserved for development purposes.
9. If the run was terminated by BARON.
10. If the run was terminated by BARON’s parser because of a syntax error in the BARON
input file.
11. If the run was terminated because of a licensing error.
12. If the heuristic termination rule was invoked by the user.

• BARON’s model status, which can take one of the following values:

1. optimal within tolerances

14 BARON user manual v. 15.6.5

2. infeasible
3. unbounded
4. intermediate feasible
5. unknown

• If model status is 4 or 5, this entry denotes the number of missing bounds from vari-
ables/expressions that make BARON unable to guarantee global optimality.

• The number of branch-and-bound iterations taken

• The node where the best solution was found (nodeopt).

• The maximum number of nodes stored in memory.

• The total CPU time in seconds.

If nodeopt = −3, there will be no solution in the results file. Otherwise, the solution can be
found in the results file by starting from the end of the file, searching backward for “***” and
then reading the solution forward, one variable at a time. The variables are ordered in the way
they were defined in the VARIABLES section of the BARON file. If available, the dual solution is
also provided there. In addition, the best primal solution is provided using variable names. If
the solution process is interrupted, for instance by Ctrl-C, the primal solution will be present in
the results file but not necessarily the corresponding dual.
If BARON declares the problem as unbounded, it will report its best solution found, possibly
followed by a vertex and direction of an unbounded ray at the end of the results file.
In the case of numsol > 1, BARON returns the best numsol solutions found. These solutions
follow right after the “***” mentioned above and they are sorted in improving order; the last
solution is the best. Duals may not be returned for all these solutions. For those solutions for
which a corresponding dual was found, the dual is also printed right after the primal. There will
typically be a dual solution for the best solution found and all local minima. However, there will
be no dual for non-KKT points, something that is highly likely to happen in most applications.
When numsol = −1 (find all feasible solutions), the solutions are reported in the results file as
soon as they are found. These solutions are reported before the “***” and can be read from this
file by searching for occurrences of “found”, reading the solution reported immediately thereafter,
and repeating this process until all occurrences of “found” are identified. Again, many of these
solutions will be reported without corresponding duals. At the end of the file, i.e., following ’‘***
Succ...”, the best solution can be read, along with a corresponding dual.

6 Some BARON features

The features described in this section rely on options that are further detailed in the next section.
For details of the algorithmic implementations, the user may wish to consult publications cited
at the end of this document.
BARON user manual v. 15.6.5 15

6.1 No starting point is required

In contrast to many NLP algorithms that require a feasible starting point, a starting point is
not required for BARON. A user may optionally provide a starting point for all or even some of
the problem variables. BARON will judiciously initialize any variables that are not initialized
by the user. Even when the problem functions cannot be evaluated at a user-provided starting
point, BARON is still capable of carrying out its global search.

6.2 Finding a few of the best or all feasible solutions

BARON offers a facility, through its NumSol option to find the best few, or even all feasible, solu-
tions to a model. The development of this facility was motivated by combinatorial optimization
problems but the facility is applicable to continuous problems as well. Even for combinatorial
problems, BARON does not rely on integer cuts to find multiple solutions. Instead, it uti-
lizes a single search tree, thus providing a computationally efficient method for finding multiple
solutions. Furthermore, because BARON’s approach applies to integer as well as continuous
programs, it can be used to find all feasible solutions to a system of nonlinear equality and
inequality constraints.
Once a model is solved by BARON with the NumSol option, the solutions found can be read from
BARON results file. To illustrate this feature, we consider a problem in kinematic analysis of
robot manipulators, the so-called indirect-position or inverse kinematics problem, in which the
desired position and orientation of a robot hand is given and the relative robot joint displacements
are to be found. The specific example that we consider involves the following set of equations
for the PUMA robot:

γ1 x1 x3 + γ2 x2 x3 + γ3 x1 + γ4 x2 + γ5 x4 + γ6 x7 + γ7 = 0
γ8 x1 x3 + γ9 x2 x3 + γ10 x1 + γ11 x2 + γ12 x4 + γ13 = 0
γ14 x6 x8 + γ15 x1 + γ16 x2 = 0
γ17 x1 + γ18 x2 + γ19 = 0
x21 + x22 − 1 = 0
x23 + x24 − 1 = 0
x25 + x26 − 1 = 0
x27 + x28 − 1 = 0
−1 ≤ xi ≤ 1, i = 1, . . . , 8

where

γ1 = 0.004731 γ6 = 1 γ11 = −0.07745 γ16 = 0.004731

γ2 = −0.3578 γ7 = −0.3571 γ12 = −0.6734 γ17 = −0.7623
γ3 = −0.1238 γ8 = 0.2238 γ13 = −0.6022 γ18 = 0.2238
γ4 = −001637 γ9 = 0.7638 γ14 =1 γ19 = 0.3461
γ5 = −0.9338 γ10 = 0.2638 γ15 = 0.3578

The first four equations of this problem are bilinear while the last four are generalized cylinders.
BARON’s scheme for finding all feasible solutions works well in continuous spaces as long as the
16 BARON user manual v. 15.6.5

solutions are isolated (separated by a certain distance). The BARON option isoltol (default
value of 10−4 ) allows the user to specify the isolation tolerance to be used to discriminate among
different solutions. In order for two feasible solution vectors to be considered different, at least
one of their coordinates must differ by isoltol.
The BARON file for the robot problem is as follows:

// Filename: robot.bar
//
// Purpose: Find all solutions of the PUMA robot problem
// L.-W. Tsai and A. P. Morgan, "Solving the kinematics of the
// most general six- and five-degree-of-freedom manipulators by
// continuation methods," ASME J. Mech. Transm. Automa. Des.,
// 107, 189-200, 1985.

OPTIONS{
numsol: 20;
}

VARIABLES x1,x2,x3,x4,x5,x6,x7,x8;

LOWER_BOUNDS{
x1: -1;
x2: -1;
x3: -1;
x4: -1;
x5: -1;
x6: -1;
x7: -1;
x8: -1;
}

UPPER_BOUNDS{
x1: 1;
x2: 1;
x3: 1;
x4: 1;
x5: 1;
x6: 1;
x7: 1;
x8: 1;
}

EQUATIONS e2,e3,e4,e5,e6,e7,e8,e9,e10,e11,e12,e13,e14,e15,e16;

e2: 0.004731*x1*x3 - 0.1238*x1 - 0.3578*x2*x3 - 0.001637*x2 - 0.9338*x4 + x7

BARON user manual v. 15.6.5 17

<= 0.3571;

e3: 0.1238*x1 - 0.004731*x1*x3 + 0.3578*x2*x3 + 0.001637*x2 + 0.9338*x4 - x7

<= -0.3571;

e4: 0.2238*x1*x3 + 0.2638*x1 + 0.7623*x2*x3 - 0.07745*x2 - 0.6734*x4 - x7

<= 0.6022;

e5: (-0.2238*x1*x3) - 0.2638*x1 - 0.7623*x2*x3 + 0.07745*x2 + 0.6734*x4 + x7

<= -0.6022;

e6: x6*x8 + 0.3578*x1 + 0.004731*x2 <= 0;

e7: - x6*x8 - 0.3578*x1 - 0.004731*x2 <= 0;

e8: - 0.7623*x1 + 0.2238*x2 == -0.3461;

e9: x1^2 + x2^2 <= 1;

e10: (-x1^2) - x2^2 <= -1;

e11: x3^2 + x4^2 <= 1;

e12: (-x3^2) - x4^2 <= -1;

e13: x5^2 + x6^2 <= 1;

e14: (-x5^2) - x6^2 <= -1;

e15: x7^2 + x8^2 <= 1;

e16: (-x7^2) - x8^2 <= -1;

OBJ: minimize 0;

The above problem has 14 different solutions. Looking at the BARON results file, all these
solutions can be found after the “*** Normal Completion ***” message.

6.3 Using BARON as a multi-start heuristic solver

To gain insight into the difficulty of a nonlinear program, especially with regard to existence of
multiple local solutions, modelers often make use of multiple local searches from randomly gen-
erated starting points. This can be easily done with BARON’s NumLoc option, which determines
the number of local searches to be done by BARON’s preprocessor. BARON can be forced to
terminate after preprocessing by setting the number of iterations to 0 through the MaxIter op-
tion. In addition to local search, BARON’s preprocessor performs extensive reduction of variable
18 BARON user manual v. 15.6.5

ranges. To sample the search space for local minima without range reduction, one would have
to set to 0 the range reduction options TDo, MDo, LBTTDo, and OBTTDo. On the other hand,
leaving these options to their default values increases the likelihood of finding high quality local
optima during preprocessing. If NumLoc is set to −1, local searches in preprocessing will be done
from randomly generated starting points until global optimality is proved or MaxTime seconds
have elapsed.

6.4 Systematic treatment of unbounded problems

If BARON declares a problem as unbounded, it will search for and may report a vertex and
direction of an unbounded ray. In addition, BARON will report the best solution found. This
will be a feasible point that is as far as possible along an unbounded ray while avoiding numerical
errors due to floating point arithmetic.

6.5 Systematic treatment of infeasible problems

If BARON declares a problem as infeasible, it has the capability to identify a subset of the
constraints that are infeasible and become feasible once any one of them is eliminated. This,
so-called, irreducibly inconsistent system (IIS) can be obtained by BARON for all types of prob-
lems handled by BARON, including linear and nonlinear, continuous and integer, convex and
nonconvex, and problems with complementarity constraints. BARON’s CompIIS option can be
used to identify an IIS.
As an example, consider the problem of minimizing the nonconvex function x1 x3 over the fol-
lowing nonconvex constrained set:

e1 : 85 + 0.006x2 x5 + 0.0006x1 x4 − 0.002x3 x5 <= 92

e2 : 0.8x2 x5 + 0.003x1 x2 + 0.002x23 = 110
e3 : 9 + 0.005x3 x5 + 0.001x1 x3 + 0.002x3 x4 <= 25
78 ≤ x1 ≤ 102
33 ≤ x2 ≤ 45
27 ≤ xi ≤ 45, i = 3, . . . , 5

When this problem is solved with CompIIS equal to 1, BARON provides the following result:

IIS contains 1 row and 3 columns as follows:

e2 Upper
x1 Lower
x2 Lower
x5 Lower

The IIS consists of the lower bounds of variables x1 , x2 and x5 , along with the ≤ part of
the equality constraint e2. This suggests that constraint e2 and the entire model can be made
feasible by lowering the lower bound of any of the three variables that are part of the IIS, whereas
modifying the bounds of x3 would not make the model feasible.
BARON user manual v. 15.6.5 19

If a problem is known to be infeasible and the user desires to identify an IIS, it may be beneficial
to set BARON’s NumLoc option to zero. Doing so will deactivate BARON’s initial upper bounding
search, which involves multiple local searches. On the other hand, DoLocal should be nonzero
in order to permit local search during the solution of certain subproblems that BARON solves
while searching for an IIS.

6.6 Handling of complementarity constraints

Complementarity relationships of the type f (x)g(x) = 0 are automatically recognized and ex-
ploited algorithmically by BARON. The functions f and g may be univariate or multivariate,
linear or nonlinear, convex or nonconvex, in terms of continuous and/or integer variables, and
may be subject to additional constraints in the model. These complementarity relationships can
be inferred by BARON even when implied by problem constraints and variable bounds. As a
result, BARON can solve general mathematical programs with equilibrium constraints (MPECs).
This class of problems includes the classical linear complementarity problem

(LCP): Find z ≥ 0 and q such that Mz + q ≥ 0 and z t (Mz + q) = 0

as well as the more general mixed complementarity problem

(MCP): Given a function f : Rn → Rn and bounds l, u ∈ Rn with

R = R ∪ {−∞, +∞}, find z ∈ Rn and w, v ∈ Rn+ such that
f (z) = w − v, l ≤ z ≤ u, (z − l)t w = 0, (u − z)t v = 0

Both problems are automatically recognized and exploited by BARON without the user having
to mark complementarities in any special way.

6.7 Parallel capabilities

For difficult problems with integer variables, most of BARON’s time is spent on solving MIP
relaxations. Hence, considerable speedups may be obtained via parallel solution of the MIP
subproblems. For this purpose, the option threads may be used to specify the number of cores
that BARON’s MIP subsolver is allowed to use. By default, this option has the value of 1,
meaning that a single core will be utilized.

7 The BARON options

The BARON options allow the user to control termination tolerances, branching and relaxation
strategies, heuristic local search options, and output options as detailed in this section.
Contrary to variable names, the BARON parser is not case-sensitive to option names.

7.1 Termination options

20 BARON user manual v. 15.6.5

Option Description Default

EpsA (ǫa ) Absolute termination tolerance. BARON terminates if 1e-6
|U − L| ≤ ǫa , where U and L are the values of the in-
cumbent and best estimate, respectively, for the optimiza-
tion problem at the current iteration. EpsA must be a real
greater than or equal to 1e-12.
EpsR (ǫr ) Relative termination tolerance. BARON terminates if 1e-4
|U − L| ≤ ǫr |L|, where U and L are the values of the
incumbent and best estimate, respectively, for the opti-
mization problem at the current iteration. EpsR must be
a nonnegative real.
DeltaTerm Users have the option to request BARON to terminate if 0
insufficient progress is made over δt consecutive seconds.
Progress is measured using the absolute and relative im-
provement thresholds δa and δr defined below. Termi-
nation will occur if, over a period of δt consecutive sec-
onds, the value of the best solution found by BARON is
not improved by at least an absolute amount δa or an
amount equal to δr times the value of the incumbent at
time t − δt . This termination condition is enforced after
processing the root node and only after a feasible solu-
tion has been obtained. Because it relies on CPU time
measurements, which may depend on machine load, this
option may result in nondeterministic behavior.
0: do not enforce this termination condition
1: terminate if progress is insufficient
DeltaT (δt ) If DeltaTerm is set to 1, BARON will terminate if insuffi- -100
cient progress is made over δt consecutive seconds. If δt is
set to a non-positive quantity, BARON will automatically
set δt equal to −δt times the CPU time taken till the end
of root node processing. DeltaT can take any real value.
DeltaA (δa ) Absolute improvement termination threshold. DeltaA ∞
must be a real greater than or equal to 1e-12.
DeltaR (δr ) Relative improvement termination threshold. DeltaR 1
must be a real greater than or equal to 1e-12.
CutOff BARON may ignore parts of the search space that contain ∞
solutions that are no better than this value. CutOff can
take any real value.
AbsConFeasTol Absolute constraint feasibility tolerance. This tolerance is 1e-5
used for general constraints and variable bounds. A point
is considered feasible for a constrain/bound if the abso-
lute or relative constraint feasibility tolerance is satisfied.
AbsConFeasTol must be a real greater than or equal to
1e-12.
BARON user manual v. 15.6.5 21

RelConFeasTol Relative constraint feasibility tolerance. This tolerance is 0

used for general constraints and variable bounds. A point
is considered feasible for a constrain/bound if the abso-
lute or relative constraint feasibility tolerance is satisfied.
RelConFeasTol must be a real between 0 and 0.1.
AbsIntFeasTol Absolute integer feasibility tolerance. All integer variable 1e-5
values must satisfy this tolerance. A point is considered
integer feasible for a variable if integrality is satisfied us-
ing the absolute or relative integer feasibility tolerance.
AbsIntFeasTol must be a real greater than or equal to
1e-12.
RelIntFeasTol Relative integer feasibility tolerance. All integer variable 0
values must satisfy this tolerance. A point is considered
integer feasible for a variable if integrality is satisfied us-
ing the absolute or relative integer feasibility tolerance.
RelIntFeasTol must be a real between 0 and 0.1
BoxTol Boxes will be eliminated if smaller than this tolerance. 1e-8
BoxTol must be a real greater than or equal to 1e-12.
FirstFeas If set to 1, BARON will terminate once it finds NumSol 0
feasible solutions, irrespective of solution quality. By de-
fault, FirstFeas is 0, meaning that BARON will search
for the best NumSol feasible solutions.
0: do not enforce this termination condition
1: terminate as soon as NumSol feasible solutions
are found
MaxIter Maximum number of branch-and-reduce iterations al- -1
lowed. −1 implies unlimited. Setting MaxIter to 0 will
force BARON to terminate after root node preprocessing.
Setting MaxIter to 1 will result in termination after the
solution of the root node. MaxIter must be an integer
greater than or equal to −1.
MaxTime Maximum time allowed (sec). For single-threaded jobs, 1000
i.e., when threads equals 1, this limit is enforced on CPU
time consumed by the job. For multi-threaded jobs, the
MaxTime limit is enforced on wall clock time. Setting
MaxTime to −1 will make BARON ignore the time limit.
MaxTime must be a real equal to −1 or greater than 0.
NumSol Number of feasible solutions to be found. Solutions found 1
will be listed in the results file. As long as NumSol 6=
-1, these solutions will be sorted from best to worse. If
NumSol is set to −1, BARON will search for all feasible
solutions to the given model and print them, in the order
in which they are found, in the results file. NumSol must
be an integer greater than or equal to −1.
22 BARON user manual v. 15.6.5

IsolTol Separation distance between solutions. This option is used 1e-4

in conjunction with NumSol. For combinatorial optimiza-
tion problems, feasible solutions are isolated. For contin-
uous problems, feasible solution points within an l∞ dis-
tance that does not exceed IsolTol > 0 will be treated
as identical by BARON. IsolTol can take any positive
value greater than or equal to 1e-12.

7.2 Relaxation options

Option Description Default

NOuter1 Number of outer approximators of convex univariate func- 4
tions. NOuter1 must be a nonnegative integer.
NOutPerVar Number of outer approximators per variable for convex 4
multivariate functions. NOutPerVar must be a nonnega-
tive integer.
NOutIter Number of rounds of cutting plane generation at node re- 4
laxation. NOutIter must be a nonnegative integer.
OutGrid Number of grid points per variable for convex multivariate 20
approximators of BARON’s CONVEX EQUATIONS. OutGrid
must be a nonnegative integer.

7.3 Range reduction options

Option Description Default

TDo Nonlinear-feasibility-based range reduction option (poor 1
man’s NLPs).
0: no bounds tightening is performed
1: bounds tightening is performed
MDo Marginals-based reduction option. 1
0: no range reduction based on marginals
1: range reduction done based on marginals
LBTTDo Linear-feasibility-based range reduction option (poor 1
man’s LPs).
0: no range reduction based on feasibility
1: range reduction done based on feasibility
OBTTDo Optimality-based tightening option. 1
0: no range reduction based on optimality
1: range reduction done based on optimality
PDo Number of probing problems allowed. -2
-2: automatically decided by BARON
0: no range reduction by probing
-1: probing on all variables
n: probing on n variables
BARON user manual v. 15.6.5 23

7.4 Tree management options

Option Description Default

BrVarStra Branching variable selection strategy. 0
0: BARON’s dynamic strategy
1: largest violation
2: longest edge
BrPtStra Branching point selection strategy. 0
0: BARON’s dynamic strategy
1: ω-branching
2: bisection-branching
3: convex combination of ω and bisection
NodeSel Specifies the node selection rule to be used for exploring 0
the search tree.
0: BARON’s strategy
1: best bound
2: LIFO
3: minimum infeasibilities

7.5 Local search options

Option Description Default

DoLocal Local search option for upper bounding. 1
0: no local search is done during upper bounding
1: BARON automatically decides when to apply lo-
cal search based on analyzing the results of pre-
vious local searches
NumLoc Number of local searches done in preprocessing. The first −2
one begins with the user-specified starting point. Sub-
sequent local searches are done from judiciously chosen
starting points. If NumLoc is set to −1, local searches
in preprocessing will be done until proof of globality or
MaxTime is reached. If NumLoc is set to −2, BARON de-
cides the number of local searches in preprocessing based
on problem and NLP solver characteristics. NumLoc must
be an integer greater than or equal to −2.

7.6 Output options

Option Description Default

PrFreq Log output frequency in number of nodes. 1000000
PrTimeFreq Log output frequency in number of seconds. 30
24 BARON user manual v. 15.6.5

PrLevel Option to control log output. 1

0: all log output is suppressed
1: print log output
LocRes Option to control output to log from local search. 0
0: no local search output
1: detailed results from local search will be printed
to the results file
ProName Problem name. This option must be provided in double problem
quotes and be no longer than 10 characters.
results Indicator if a results file is to be created. 1
0: do not create file
1: create file named according to the ResName op-
tion
ResName Name of results file to be written. This option must be res.lst
provided in double quotes in the .bar file.
summary Indicator if a summary file is to be created. 0
0: do not create file
1: create file named according to the SumName op-
tion.
SumName Name of summary file to be written. This option must be sum.lst
provided in double quotes in the .bar file.
times Indicator if a times file is to be created. 0
0: do not create file
1: create file named according to the TimName op-
tion.
TimName Name of times file to be written. This option must be tim.lst
provided in double quotes in the .bar file.

7.7 Subsolver options

Option Description Default

LPSol Specifies the LP/MIP solver to be used. 3
3: CPLEX
8: CLP/CBC
BARON user manual v. 15.6.5 25

libcplex.so
CplexLibName If utilized, this option must be supplied in double quotes cplex.dll
and provide the entire path to the location of the CPLEX libcplex.dylib
callable libraries on the user’s computer. If left unspecified
and LPSol is 3, BARON will utilize standard library loca-
tion facilities to locate CPLEX and use it for the solution
of LP/NLP subproblems. In the latter case, the CPLEX
libraries should be in the user’s LIBRARY PATH. When
searching for the libraries, on Windows systems, BARON
will look for cplex.dll, whereas on Linux systems it will
look for libcplex.so; on MAC OSX, it will look for libc-
plex.dylib. If a CPLEX library named CplexLibName is
not found, BARON will search for a library containing the
version number (currently cplex1261.dll, libcplex1261.so
or libcplex1261.dylib, depending on the platform). If
CPLEX is still not found, BARON will resort to using
CLP/CBC instead. The CplexLibName option is not ap-
plicable to BARON under AIMMS, AMPL or GAMS. As
an alternative to the CplexLibName option, users may
copy the CPLEX libraries or use symbolic links to place
the default library name on their user LIBRARY PATH.
LPAlg Specifies the LP algorithm to be used. 0
0: automatic selection of LP algorithm
1: primal simplex
2: dual simplex
3: barrier
NLPSol Specifies the NLP solver to be used. By default, BARON -1
will select the NLP solver and may switch between dif-
ferent NLP solvers during the search, based on problem
characteristics and solver performance. Any combination
of licensed NLP solvers may be used in that case. A single
specific NLP solver can be specified by setting this option
to a value other than the default. If the specified solver
is not licensed, BARON will default to automatic solver
selection.
-1: automatic solver selection
0: Local search based on function evaluations alone
with no calls to local solvers
9: IPOPT
10: FilterSD
AllowFilterSD In case of automatic NLP solver selection, this option can 0
be used to selectively permit or disallow the use of Fil-
terSD as an NLP subsolver.
0: do not use FilterSD for local search
1: consider FilterSD for local search
26 BARON user manual v. 15.6.5

AllowIpopt In case of automatic NLP solver selection, this option can 1

be used to selectively permit or disallow the use of IPOPT
as an NLP subsolver.
0: do not use IPOPT for local search
1: consider IPOPT for local search

7.8 Licensing options

Option Description Default

LicName License file name. If utilized, this option must be supplied baronlice.txt
in double quotes and provide the entire path to the loca-
tion of the BARON license file. If left unspecified, BARON
will search for the BARON license file baronlice.txt in
the user PATH. This option is not applicable to BARON
under AIMMS, AMPL or GAMS.

7.9 Other options

Option Description Default

CompIIS In case of an infeasible problem, this option can be used 0
to search for an IIS. Setting this option equal to 1, works
very well for most problems.
0: do not search for an IIS
1: the search for an IIS is based on a fast heuristic
2: an IIS is obtained using a deletion filtering algo-
rithm
3: an IIS is obtained using an addition filtering al-
gorithm
4: an IIS is obtained using an addition-deletion fil-
tering algorithm
5: an IIS is obtained using a depth-first search al-
gorithm
BARON user manual v. 15.6.5 27

IISint When search for an IIS is requested through CompIIS, 0

BARON assumes that the model is unlikely to include an
error in terms of binaries, i.e. the binary definitions are
assumed correct and the IIS output should be interpreted
with respect to binary definitions. General integer bounds
may be assumed as correct or can be questioned using the
option IISint, which can take the values of 0 (default,
where integer bounds are assumed correct and the IIS
should be interpreted with respect to integer bounds) or
1 (signalling that general integer bounds should be ques-
tioned). Integrality is enforced in both cases.
0: do not consider general integers as part of an IIS
1: consider general integers (but not binaries) as
part of an IIS
IISorder Defines the order in which problem constraints are consid- -1
ered in the search for an IIS.
-1: auto set to aim for a small IIS depending on the
value of CompIIS
1: arrange constraints in problem order
2: arrange constraints in ascending order of degree
3: arrange constraints in descending order of degree
>=4: random order using IISorder as seed
threads Specifies the number of cores that BARON is allowed 1
to use for solution of its MIP subproblems. By de-
fault, this option has the value of 1, meaning that a
single core will be utilized. The value of this option is
passed to CBC, CPLEX, and XPRESS through the op-
tions CPX PARAM THREADS, XPRS THREADS, and
-threads, respectively.

8 Bibliography
The following is a partial list of BARON-related publications that describe the algorithms im-
plemented in the software, the theory behind them, and some related applications.

1. H. S. Ryoo and N. V. Sahinidis. Global optimization of nonconvex NLPs and MINLPs with
applications in process design. Computers & Chemical Engineering, 19:551–566, 1995.
2. M. C. Dorneich and N. V. Sahinidis. Global optimization algorithms for chip layout and
compaction. Engineering Optimization, 25:131–154, 1995.
3. H. S. Ryoo and N. V. Sahinidis. A branch-and-reduce approach to global optimization.
Journal of Global Optimization, 8:107–139, 1996.
4. N. V. Sahinidis. BARON: A general purpose global optimization software package. Journal
of Global Optimization, 8:201–205, 1996.
28 BARON user manual v. 15.6.5

5. R. A. Gutierrez and N. V. Sahinidis. A branch-and-bound approach for machine selection

in just-in-time manufacturing systems. International Journal of Production Research,
34:797–818, 1996.

6. M. L. Liu, N. V. Sahinidis, and J. P. Shectman. Planning of chemical process networks via

global concave minimization. In I. E. Grossmann (ed.), Global Optimization in Engineering
Design, Kluwer Academic Publishers, Boston, MA, pages 195–230, 1996.

7. V. Ghildyal. Design and Development of a Global Optimization System. Master’s thesis,

Department of Mechanical & Industrial Engineering, University of Illinois, Urbana, IL,
1997.

8. J. G. VanAntwerp, R. D. Braatz, and N. V. Sahinidis. Globally optimal robust control for

systems with nonlinear time-varying perturbations. Computers & Chemical Engineering,
21:S125–S130, 1997.

9. J. P. Shectman and N. V. Sahinidis. A finite algorithm for global minimization of separable

concave programs. Journal of Global Optimization, 12:1–36, 1998.

10. J. G. VanAntwerp, R. D. Braatz, and N. V. Sahinidis. Globally optimal robust control. J.

Process Control, 9:375–383, 1999.

11. N. V. Sahinidis and M. Tawarmalani. Applications of global optimization to process and

molecular design. Computers & Chemical Engineering, 24:2157–2169, 2000.

12. V. Ghildyal and N. V. Sahinidis. Solving global optimization problems with BARON.
In A. Migdalas, P. Pardalos, and P. Varbrand (eds.), From Local to Global Optimization.
A Workshop on the Occasion of the 70th Birthday of Professor Hoang Tuy, Linköping,
Sweden, Aug. 24-29, 1997, Kluwer Academic Publishers, Boston, MA, pages 205–230,
2001.

13. H. S. Ryoo and N. V. Sahinidis. Analysis of bounds for multilinear functions. Journal of
Global Optimization, 19:403–424, 2001.

14. M. Tawarmalani and N. V. Sahinidis. Semidefinite relaxations of fractional programs

via novel techniques for constructing convex envelopes of nonlinear functions. Journal of
Global Optimization, 20:137–158, 2001.

15. M. Tawarmalani and N. V. Sahinidis. Convex extensions and convex envelopes of l.s.c.
functions. Mathematical Programming, 93:247–263, 2002.

16. M. Tawarmalani and N. V. Sahinidis. Convexification and Global Optimization in Con-

tinuous and Mixed-Integer Nonlinear Programming: Theory, Algorithms, Software, and
Applications. Kluwer Academic Publishers, Dordrecht, 2002.

17. M. Tawarmalani, S. Ahmed, and N. V. Sahinidis. Global optimization of 0−1 hyperbolic

programs. Journal of Global Optimization, 24:385–417, 2002.

18. M. Tawarmalani, S. Ahmed, and N. V. Sahinidis. Product disaggregation and relaxations

of mixed-integer rational programs. Optimization and Engineering, 3:281–303, 2002.
BARON user manual v. 15.6.5 29

19. N. V. Sahinidis, M. Tawarmalani, and M. Yu. Design of alternative refrigerants via global
optimization. AIChE Journal, 49:1761–1775, 2003.

20. H. S. Ryoo and N. V. Sahinidis. Global optimization of multiplicative programs. Journal

of Global Optimization, 26:387–418, 2003.

21. N. V. Sahinidis. Global optimization and constraint satisfaction: The branch-and-reduce

approach. In AC. Bliek, C. Jermann, and A. Neumaier (eds.), Global Optimization and
Constraint Satisfaction, Lecture Notes in Computer Science, Vol. 2861, Springer, Berlin,
pages 1–16, 2003.

22. S. Ahmed, M. Tawarmalani, and N. V. Sahinidis. A finite branch-and-bound algorithm for

two-stage stochastic integer programs. Mathematical Programming, 100:355–377, 2004.

23. M. Tawarmalani and N. V. Sahinidis. Global optimization of mixed-integer nonlinear

programs: A theoretical and computational study. Mathematical Programming, 99:563–
591, 2004.

24. N. V. Sahinidis and M. Tawarmalani. Accelerating branch-and-bound through a modeling

language construct for relaxation-specific constraints. Journal of Global Optimization,
32:259–280, 2005.

25. M. Tawarmalani and N. V. Sahinidis. A polyhedral branch-and-cut approach to global

optimization. Mathematical Programming, 103:225–249, 2005.

26. N. V. Sahinidis. BARON: A general purpose global optimization software package. Journal
of Global Optimization, 8:201–205, 1996.

27. Y. Chang and N. V. Sahinidis. Global optimization in stabilizing controller design. Journal
of Global Optimization, 38:509–526, 2007.

28. X. Bao, N. V. Sahinidis, and M. Tawarmalani. Multiterm polyhedral relaxations for non-
convex, quadratically-constrained quadratic programs. Optimization Methods and Soft-
ware, 24:485–504, 2009.

29. L. M. Rios and N. V. Sahinidis. Portfolio optimization for wealth-dependent risk prefer-
ences. Annals of Operations Research, 177:63–90, 2010.

30. X. Bao, N. V. Sahinidis, and M. Tawarmalani. Semidefinite relaxations for quadratically

constrained quadratic programming: A review and comparisons. Mathematical Program-
ming, 129:129–157, 2011.

31. A. Khajavirad and N. V. Sahinidis. Convex envelopes of products of convex and component-
wise concave functions. Journal of Global Optimization, 52:391–409, 2012.

32. A. Khajavirad and N. V. Sahinidis. Convex envelopes generated from finitely many com-
pact convex sets. Mathematical Programming, 137:371–408, 2013.

33. K. Zorn and N. V. Sahinidis. Global optimization of general nonconvex problems with
intermediate bilinear substructures. Optimization Methods and Software, 29:442–462, 2013.
30 BARON user manual v. 15.6.5

34. K. Zorn and N. V. Sahinidis. Computational experience with applications of bilinear

cutting planes. Industrial & Engineering Chemistry Research, 52:7514-7525, 2013.

35. A. Khajavirad, J. J. Michalek and N. V. Sahinidis. Relaxations of factorable functions

with convex-transformable intermediates. Mathematical Programming, 144:107-140, 2014.

36. K. Zorn and N. V. Sahinidis. Global optimization of general nonconvex problems with inter-
mediate polynomial substructures. Journal of Global Optimization, DOI: 10.1007/s10898-
014-0190-2, 2014.

37. X. Bao, A. Khajavirad, N. V. Sahinidis, and M. Tawarmalani. Global optimization of non-

convex problems with multilinear intermediates. Mathematical Programming Computation,
DOI: 10.1007/s12532-014-0073-z, 2015.