Journal of Theoretical and Applied Information Technology
20
th
October 2013. Vol. 56 No.2
© 2005 - 2013 JATIT & LLS. All rights reserved
.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
362
CONTINUOUS HOPFIELD NETWORK AND QUADRATIC
PROGRAMMING FOR SOLVING THE BINARY
CONSTRAINT SATISFACTION PROBLEMS
1
KHALID HADDOUCH,
1*
MOHAMED ETTAOUIL and
2
CHAKIR LOQMAN
1
UFR: Scientific Computing and Computer sciences, Engineering sciences, Faculty of Science
and Technology, University Sidi Mohamed Ben Abdellah Box 2202, Fez, Morocco
2
Department of Computer Engineering, Moulay Ismail University, High School of technology, B.
P. 3103, 50000, Toulal, Meknes, Morocco
E-mail
: 1
haddouchk@yahoo.fr ,
1*
mohamedettaouil@yahoo.fr ,
2
chakirfst@yahoo.fr
ABSTRACT
Many important computational problems may be formulated as constraint satisfaction problems (CSP). In
this paper, we propose a new approach to solve the binary CSP problems using the continuous Hopfield
networks (CHN). This approach is divided into three steps: the first concerns reducing the size of the CSP
problems using arc consistency technique AC3. The second step involves modeling the filtered constraint
satisfaction problems as 0-1 quadratic programming subject to linear constraints. The last step concerns
applying the continuous Hopfield networks to solve the obtained 0-1 optimization model. Therefore, the
mapping procedure and an appropriate parameter setting procedure about CSP problems are given in detail.
Finally, some computational experiments solving the CSP problems are shown.
Keywords: Constraint Satisfaction Problems, Filtering Algorithms, Quadratic 0-1 Programming,
Continuous Hopfield Networks, Energy Function.
1. INTRODUCTION
A large number of problems in artificial
intelligence and other areas of computer science can
be viewed as special cases of the constraint
satisfaction problems. Some examples include
machine vision, belief maintenance, scheduling,
temporal reasoning, graph problems, aircraft
conflict, etc.
A constraint satisfaction problem is defined by a
set of variables, a finite and discrete domain for
each variable and a set of constraints. Each
constraint restricts the combination of values that a
set of variables may take simultaneously. Solving
the CSP problem requires assigning a value for each
variable from each domain in such a way that all
constraints are satisfied. The task is to find one
solution or all solutions. However, the CSP
problems are NP-complete problems requiring a
combination of heuristics and combinatorial search
methods in order to be solved in a reasonable
time[12].
A number of different approaches have been
developed for solving the constraint satisfaction
problems [10], [18], [19], [21]. Some of them use
backtracking to directly search for possible
solutions [5], others use consistency techniques to
simplify the original problems [3], [4], [24], and
some are a combination of these two techniques [2],
[5], [13]. Moreover, we proposed different
approaches to solve the constraint satisfaction
problems. The first one consists of modeling a
constraint satisfaction problem as 0-1 quadratic
knapsack problem subject to quadratic constraint
[6], [7]. The second one is a new model of the
binary CSP problem as 0-1 quadratic programming
which consists in minimizing the quadratic function
subject to linear constraints[8]. The later model will
be used in order to determine a generalized energy
function for the binary constraint satisfaction
problem. This step is the most important one for
solving any binary CSP problem using continuous
Hopfield networks.
This neural network was introduced by Hopfield
and Tank [13], [14] and it has been extensively
studied, developed and has found many applications
in many areas, such as pattern recognition, model
identification, and optimization [11], [28].
Moreover, this neural network is applied to solve
many problems such as traveling salesman
problems [27], graph coloring problems [17],
allocation problems [16], etc. It is important to note