Improved Stability Criteria of Generalized Recurrent
Neural Networks with Time-varying Delays
Jinfang Zhang1 Yuanhua Qiao1 Jun Miao2 Lijuan Duan3
1
College of Applied Science, Beijing University of Technology,
Beijing, 100124, China
2
Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China
3
College of Computer Science, Beijing University of Technology, Beijing 100124, China
Abstract - In this paper, the problem on stability analysis of
generalized recurrent neural networks with a time-varying delays is
considered. Neither the differentiability, the monotony on these
activation functions nor the differentiability on the time-varying
delays are assumed. By employing a new Lyapunov-Krasovskii
function, a linear matrix inequality (LMI) approach is developed to
establish sufficient conditions for RNNs to be globally asymptotically
stable. The proposed stability results are less conservative than some
recently known ones in the literature. Finally an example is given to
verify the effectiveness of the present criterion.
Index Terms: Recurrent neural networks; Global asymptotically
stability; Time-varying delays; Linear matrix inequality
1. Introduction
It is well known that delayed neural networks have a wide
range of applications in many fields such as pattern
recognition, image processing, associative memory, and
optimization problems [1–4]. Many interesting results on
global asymptotic stability and global exponential stability
have been obtained in recent years. It is worth mentioning that
the obtained results in [5–12] are based on the following
assumptions:
the involved time-delays are constant delays
[5–8] or with time-varying delays terms but continuously
differentiable [9-12], and
the activation functions are
monotony or differentiability [13]. However, in many
situations, time delays occur frequently and vary in an
irregular fashion, and sometimes, they may be not
continuously differentiable. In such case, those results can not
be applied. The purpose of this works is to improve and
complement the results in [5–13] and present a new criterion
concerning, the global asymptotic stability of recurrent neural
networks with time-varying delays, which is independent of
the time-varying delays and does not require the
differentiability of delay functions. An example is given to
verify the effectiveness of the present criterion.
Notations. The notations are quite standard. Throughout
this paper,
and
denote, respectively,
the
dimensional Euclidean space and the set of all
real
matrices. The superscript
denotes matrix transposition. The
notation
(respectively,
) means that
and
are
symmetric matrices, and that
is positive semi-definite
(respectively, positive definite). If
is a matrix,
(respectively,
) means the largest (respectively, smallest)
eigenvalue of
. Sometimes, the arguments of a function or a
matrix will be omitted in the analysis when no confusion
arises.
2. System description
In this paper, we consider the following model:
JtdtxBftxAftCxtx )))((())(()()(
(1)
where
corresponds to the number of units in a neural
network
nT
n
Rtxtxtxtx ))(,),(),(()(
21
is the neuron state
vector;
,)))((,)),(()),((())((
2211
T
nn
txftxftxftxf
.))))(((,)),((())),(((()))(((
222111
T
nnn
tdtxftdtxftdtxftdtxf
is activation function;
corresponds to the
transmission delay and satisfies
(
is a
constant);
represents the rate
constant with which the
th unit will reset its potential to the
resting state in isolation when it is disconnected from the
networks and without external inputs;
is referred to
as the feedback matrix,
represents the delayed
feedback matrix.
is the constant external
input vector. The initial condition of model (1) is of the
form
where
is bounded and continuous on
.
Moreover, the neuron activation functions satisfy the
following assumption:
(H)They are assumed to bounded and there exist two diagonal
matrices
and
such that
(2)
for all
To prove our main results, we need the following lemmas:
Lemma 1 [14].Given constant matrices
and
, where
, then the following LMI:
International Conference on Computer, Networks and Communication Engineering (ICCNCE 2013)
© 2013. The authors - Published by Atlantis Press