Model Construction of Mix-valued Logical Network
via Observed Data
Hao Chen
Department of Mathematics
Tongji University
Shanghai, China
Email: 1eli
chen@tongji.edu.cn
Jitao Sun*
Department of Mathematics
Tongji University
Shanghai, China
Email: sunjt@sh163.net
Abstract—In the present paper, the model construction prob-
lem of mix-valued logical network is investigated. By using the
semi-tensor product, we propose a new method to build a model
for mix-valued logical network via observed data. By considering
the network graph, the number of observed data required is
reduced tremendously. Finally, examples are given to illustrate
the main results.
I. INTRODUCTION
The Boolean network was introduced by Kauffman in [1] to
describe the genetic circuits and then it is widely investigated
in different fields, such as biology, system science, physics and
so on; see [2], [3], [4] and the references therein.
A model verified by observed data is called a realization
of the data. The model construction problem is also called the
identification problem. Various algorithms have been proposed
for the network identification. A general identification algo-
rithm was proposed for the identification of genetic network
architecture in [5]. A method which required less average time
to identify randomized network was given in [6]. In [7], the
number of experiments data required to identify the network
was investigated.
Recently, the semi-tensor product (STP) of matrices was
proposed in [8] to investigate the Boolean network. It con-
verts the logical systems into discrete-time systems which is
classical [9]. The semi-tensor product (STP) of matrices has
been successfully applied in expressing and analyzing Boolean
networks. It has been applied to present many properties
of boolean networks, such as the stability and stabilization
[10], [11], the controllability and observability [12], [13], the
realization of Boolean control networks [14].
The STP method is also used in identifying the Boolean
network in [15]. It provides a new method to model a Boolean
network via observed data. It is an interesting topic in the
area of system biology. Inspiring by this, the mix-valued
network is considered in this paper. It is a challenging and
practically problem. For example, in game theory, when the
infinitely repeated game is considered, the dynamics of the
strategies depending on one history, may be expressed in mix-
valued logical network. The mix-valued networks is a general
dynamical networks. The Boolean network investigated in [15]
is a special case. Using the semi-tensor product, a new method
is proposed to build a model for mix-valued logical network
in this paper. The number of experimental data necessary is
proposed and it may be reduced tremendously if the network
graph is known.
The rest of the paper are recognized as follows. Section
2 introduces some fundamental definitions and some notations
used in the paper. In section 3, the method to build a model for
mix-valued networks via observed data is given. If the network
graph is known, then the number of experimental data required
could be reduced tremendously. In section 4, the main results
are illustrated through examples. Finally, concluding remarks
are given in section 5.
II. P
RELIMINARIES
The Semi-tensor product is the fundamental tool in the
present paper and the matrix product is assumed to be the
semi-tensor product in the following discussion. The symbol
is omitted in most cases. A review of mix-valued logic [16]
will be given.
Let
D
k
i
= {1=T,
k
i
− 2
k
i
− 1
, ...,
1
k
i
− 1
, 0=F },
which is the set from which k
i
-valued logical variables take
their values.
Δ
k
i
is used for vector expression as
Δ
k
i
= {δ
j
k
i
|j =1, 2, ..., k
i
},
where δ
j
k
i
denotes the jth column of the identity matrix I
k
i
.
For k
i
-valued logical networks, identify
j
k
i
− 1
∼ δ
k
i
−j
k
i
,j =0, 1, ..., k
i
− 1.
Let M
n×s
denote the set of n×s matrices. Assume a matrix
M =[δ
j
1
n
δ
j
2
n
... δ
j
s
n
] ∈M
n×s
, i.e., its columns, Col(M ) ⊂
Δ
n
. M is called a logical matrix. The set of n × m logical
matrices is denoted by L
n×m
.
Let X be a row vector of dimension np, and Y =
[y
1
,y
2
, ..., y
p
]
T
be a column vector of dimension p. Then we
split X into p equal-size blocks as X
1
, ..., X
p
, which are 1×n
rows. Define the semi-tensor product, denoted by ,as
⎧
⎪
⎪
⎨
⎪
⎪
⎩
X Y =
p
i=1
X
i
y
i
∈ R
n
,
Y
T
X
T
=
p
i=1
y
i
(X
i
)
T
∈ R
n
.
2012 12th International Conference on Control, Automation, Robotics & Vision
Guan
zhou, China, 5-7th December 2012 (ICARCV 2012)
978-1-4673-1872-3/12/$31.00 ©2012 IEEE
Th11.3
413