Table of Contents
ISRN Applied Mathematics
Volume 2011, Article ID 636732, 6 pages
http://dx.doi.org/10.5402/2011/636732
Research Article

Hopf Bifurcation Analysis in a Tabu Learning Neuron Model with Two Delays

School of Mathematics and Computer Science, Fujian Normal University, Fuzhou, Fujian 350007, China

Received 10 March 2011; Accepted 29 March 2011

Academic Editors: C.-C. Chiu and A. El-Sayed

Copyright © 2011 Yingguo Li. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

We consider the nonlinear dynamical behavior of a tabu leaning neuron model with two delays. By choosing the sum of the two delays as a bifurcation parameter, we prove that Hopf bifurcation occurs in the neuron. Some numerical examples are also presented to verify the theoretical analysis.

1. Introduction

Starting with the work of Hopfield [1] on neural networks, the dynamical behaviors (including stability, instability, periodic oscillatory, bifurcation, and chaos) of the continuous-time neural networks have received increasing interest due to their promising potential applications in many fields, such as signal processing, pattern recognition, optimization, and associative memories.

Tabu learning [2, 3] applies the concept of tabu search [47] to neural networks for solving optimization problems. By continuously increasing the energy surface in a neighborhood of the current network state, it penalizes those states that have already been visited. This enables the state trajectory to climb out of local minima while tending toward those areas that have not yet been visited, thus performing an efficient search through the problem’s solution space. Note that, unlike most existing neural-network-based methods for optimization, the goal of the tabu learning method is not to force the network state to converge to an optimal or nearly optimal solution, but rather, the network conducts an efficient search through the solution space. As pointed out in [8], it is very natural for one to ask what the state trajectory of the tabu learning neural network is like. In [8], the authors found and analyzed the Hopf bifurcation behavior in a tabu learning single-neuron model and also found chaotic behaviors in a tabu learning two-neuron model with both linear and quadratic proximity functions. In [9], the authors found and analyzed the Hopf bifurcation behavior in a discrete-time tabu learning single-neuron model.

It is well known that there exist time delays in the information processing of neurons. The delayed axonal signal transmissions in the neural network models make the dynamical behaviors become more complicated and may destabilize the stable equilibria and admit periodic oscillation, bifurcation, and chaos. Therefore, the delay is an important control parameter in living nervous system: different ranges of delays correspond to different patterns of neural activities (see, e.g., [10, 11]).

In the present paper, we study the bifurcation behavior of a tabu leaning neuron models with two delays. By choosing the sum of the two delays as a bifurcation parameter, we prove that Hopf bifurcation occurs in the neuron.

The organization of this paper is as follows. In the next section, we present the tabu learning neuron model with two delays that will be studied in this paper. In Section 3, we shall study the stability and the existence of bifurcation. To verify the theoretic analysis, numerical simulations are given in Section 4.

2. The Tabu Learning Neuron Model with Two Delays

Recently, Li et al. [8] have studied the following tabu learning single neuron model: 𝑢𝐽(𝑡)=𝑢(𝑡)+𝑎𝑓(𝑢(𝑡))+𝐽(𝑡),(𝑡)=𝛼𝐽(𝑡)𝛽𝑓(𝑢(𝑡)).(2.1)𝑢 is the state of neuron, where 𝑓() is the activation function and 𝑎 represents the strength of the connection from neuron to itself, and 𝑎 cannot be a negative value in Hopfield-type neural network. 𝐽 satisfied the second equation of system (2.1) which is a learning equation, where 𝛼 is the memory decay rate and 𝛽 is the learning rate. Both 𝛼 and 𝛽 are positive constants. In the tabu learning model, 𝛼, 𝛽 should be carefully chosen [12] for they heavily influence the behavior of system (2.1).

In order to describe the model more clearly, we introduce two time delays in model (2.1) 𝑢(𝑡)=𝑢(𝑡)+𝑎𝑓(𝑢(𝑡))+𝐽𝑡𝜏1,𝐽(𝑢𝑡)=𝛼𝐽(𝑡)𝛽𝑓𝑡𝜏2.(2.2) In the next section, we first consider the stability of equilibrium and then study the existence and conditions of bifurcation at the equilibrium by choosing 𝜏=𝜏1+𝜏2 as the bifurcation parameter for system (2.2).

3. Stability and Existence of Hopf Bifurcation

For most of the models in the literature, including the ones in [1315], the activation function 𝑓 is 𝑓(𝑢)=tan(𝑐𝑢). However, we only make the following assumption on function 𝑓:(𝐻)𝑓𝐶1(𝑅) and 𝑓(0)=0.

Clearly, (𝑢,𝐽)𝑇=(0,0)𝑇 is equilibrium of system (2.2). Linearization of (2.2) at the zero equilibrium yields 𝑢(𝑡)=𝑢(𝑡)+𝑎𝑓(0)𝑢(𝑡)+𝐽𝑡𝜏1,𝐽(𝑡)=𝛼𝐽(𝑡)𝛽𝑓(0)𝑢𝑡𝜏2,(3.1) whose characteristic equation is 𝜆2+𝛼+1𝑎𝑓(0)𝜆+𝛼𝑎𝛼𝑓+(0)𝛽𝑓𝑒(0)𝜆𝜏=0,(3.2) where 𝜏=𝜏1+𝜏2.

In [16, 17], the second-degree transcendental polynomial equation 𝜆2+𝑝𝜆+𝑟+(𝑠𝜆+𝑞)𝑒𝜆𝜏=0(3.3) has been extensively studied. In particular, we introduce the following results stated in [17] about the distributions of the roots of the characteristic (3.3).

Let (𝐻1)𝑝+𝑠>0, (𝐻2)𝑞+𝑟>0, (𝐻3)either 𝑠2𝑝2+2𝑟<0 and 𝑟2𝑞2>0 or (𝑠2𝑝2+2𝑟)2<4(𝑟2𝑞2), (𝐻4)either 𝑟2𝑞2<0 or 𝑠2𝑝2+2𝑟>0 and (𝑠2𝑝2+2𝑟)2=4(𝑟2𝑞2), (𝐻5)𝑟2𝑞2>0, 𝑠2𝑝2+2𝑟>0 and (𝑠2𝑝2+2𝑟)2>4(𝑟2𝑞2).

Lemma 3.1 (see [17]). For (3.3), we have the following (1)if (𝐻1)-(𝐻3) hold, then all roots of (3.3) have negative real parts for all 𝜏0; (2)if (𝐻1), (𝐻2), and (𝐻4) hold and 𝜏=𝜏+𝑗, then (3.3) has a pair of purely imaginary roots ±𝑖𝜔+. When𝜏=𝜏+0, then all roots of (3.3) except ±𝑖𝜔+ have negative real parts;(3)if (𝐻1), (𝐻2), and (𝐻5) hold and 𝜏=𝜏+𝑗(𝜏=𝜏𝑗, resp.), then (3.3) has a pair of purely imaginary roots ±𝑖𝜔+(±𝑖𝜔, resp.). Furthermore, when𝜏=𝜏+𝑗(𝜏=𝜏𝑗, resp.), then all roots of (3.3) except ±𝑖𝜔+(±𝑖𝜔, resp.) have negative real parts.Here, 𝜔±=22𝑠2𝑝2+2𝑟±𝑠2𝑝2+2𝑟2𝑟42𝑞2(1/2),𝜏±𝑗=1𝜔±𝑞𝜔arccos2±𝑟𝑝𝑠𝜔2±𝑠2𝜔2±+𝑞2+2𝑗𝜋𝜔±,𝑗=0,1,2,.(3.4)

From Lemma 3.1, the following theorem about stability and Hopf bifurcation of (2.2) are immediately obtained.

Theorem 3.2. For System (2.2), we have the following (1)if 𝛼+1𝑎𝑓(0)>0 and 𝛼𝑎𝛼𝑓(0)𝛽𝑓(0)>0, then the equilibrium (0,0) of the system (2.2) is asymptotically stable for all 𝜏0; (2)if 𝛼+1𝑎𝑓(0)>0, 𝛼𝑎𝛼𝑓(0)+𝛽𝑓(0)>0, and 𝛼𝑎𝛼𝑓(0)𝛽𝑓(0)<0, then (0,0) is asymptotically stable when 𝜏[0,𝜏+0) and unstable when 𝜏>𝜏+0. System (2.2) undergoes Hopf bifurcations at (0,0) when 𝜏=𝜏+𝑗.
Here 𝜏+𝑗=1𝜔+𝜔arccos2+𝑟𝛽𝑓+(0)2𝑗𝜋𝜔+𝜔,𝑗=0,1,2,,+=22𝛼21𝑎𝑓(0))2+𝛼21𝑎𝑓(0)24𝛼𝑎𝛼𝑓(0)2𝛽𝑓(0)2(1/2).(3.5)

Corollary 3.3. When we set the activation function 𝑓(𝑢)=tan(𝑢), notice that 𝑓(0)=1, then from Theorem 3.2 we get the following (1)if 0<𝑎<1 and 0<𝛽<𝛼(1𝑎), then the equilibrium (0,0) of the system (2.2) is asymptotically stable for all 𝜏0;(2)if either 0<𝑎<1 and 𝛽>𝛼(1𝑎) or 𝑎>1, 𝛼>𝑎1 and 𝛽>𝛼(𝑎1), then (0,0) is asymptotically stable when 𝜏[0,𝜏+0) and unstable when 𝜏>𝜏+0. System (2.2) undergoes Hopf bifurcations at (0,0) when 𝜏=𝜏+𝑗.
Here 𝜏+𝑗 and 𝜔+ are the same as (3.5).

4. Numerical Simulation

In this section, we give an example of system (2.2) with 𝑎=1.5, 𝛼=0.8, 𝛽=0.5 and 𝑓()=tan(). Then 𝑓(0)=0 and 𝑓(0)=1. The system (2.2) becomes 𝑢(𝑡)=𝑢(𝑡)+1.5tan(𝑢(𝑡))+𝐽𝑡𝜏1,𝐽𝑢(𝑡)=0.8𝐽(𝑡)0.5tan𝑡𝜏2.(4.1) In system (4.1), the case (2) of Corollary 3.3 is satisfied.

From the formulae (3.5), it follows that 𝜏+0=0.6. Thus, equilibrium (0,0) is stable when 𝜏[0,𝜏+0). Choose 𝜏1=0.1, 𝜏2=0.3; then 𝜏1+𝜏2<0.6. Figure 1(a)(A1)–(A3) shows that the origin is asymptotically stable. When 𝜏 passes through the critical value, 𝜏+0, (0,0) loses its stability and a Hopf bifurcation occurs. Choose 𝜏1=0.5, 𝜏2=0.2; then 𝜏1+𝜏2>0.6. Figure 1(b)(B1)–(B3) shows that a family of periodic solutions bifurcate from (0,0).

fig1
Figure 1: (a) (A1)–(A3): 𝜏1+𝜏2<0.6; (b) (B1)–(B3): 𝜏1+𝜏2>0.6.

Acknowledgment

The authors are very grateful to the referees for their valuable suggestions, which helped in improving the paper significantly.

References

  1. J. J. Hopfield, “Neurons with graded response have collective computational properties like those of two-state neurons,” Proceedings of the National Academy of Sciences of the United States of America, vol. 81, no. 10, pp. 3088–3092, 1984. View at Google Scholar · View at Scopus
  2. D. A. Beyer and R. G. Ogier, “Tabu learning: a neural network search method for solving nonconvexoptimization problems,” in Proceedings of the International Joint Conference on Neural Networks (IJCNN '91), pp. 953–961, Singapore, July 1991.
  3. C. Li, X. Liao, and J. Yu, “Tabu learning method for multiuser detection in CDMA systems,” Neurocomputing, vol. 49, pp. 411–415, 2002. View at Publisher · View at Google Scholar · View at Scopus
  4. F. Glover, “Tabu search, part I,” ORSA Journal on Computing, vol. 1, pp. 190–206, 1989. View at Google Scholar
  5. F. Glover, “Tabu search, part II,” ORSA Journal on Computing, vol. 2, pp. 4–32, 1990. View at Google Scholar
  6. C. Li, H. Xu, X. Liao, and J. Yu, “Tabu search for CNN template learning,” Neurocomputing, vol. 51, pp. 475–479, 2003. View at Publisher · View at Google Scholar · View at Scopus
  7. C. Li, X. Liao, and J. Yu, “Tabu search for fuzzy optimization and applications,” Information Sciences, vol. 158, no. 1–4, pp. 3–13, 2004. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet
  8. C. Li, G. Chen, X. Liao, and J. Yu, “Hopf bifurcation and chaos in tabu learning neuron models,” International Journal of Bifurcation and Chaos in Applied Sciences and Engineering, vol. 15, no. 8, pp. 2633–2642, 2005. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet
  9. M. Xiao and J. Cao, “Bifurcation analysis on a discrete-time tabu learning model,” Journal of Computational and Applied Mathematics, vol. 220, no. 1-2, pp. 725–738, 2008. View at Publisher · View at Google Scholar · View at Zentralblatt MATH
  10. C. Huang, Y. He, L. Huang, and Y. Zhaohui, “Hopf bifurcation analysis of two neurons with three delays,” Nonlinear Analysis: Real World Applications, vol. 8, no. 3, pp. 903–921, 2007. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet
  11. C. Huang, L. Huang, J. Feng, M. Nai, and Y. He, “Hopf bifurcation analysis for a two-neuron network with four delays,” Chaos, Solitons and Fractals, vol. 34, no. 3, pp. 795–812, 2007. View at Publisher · View at Google Scholar · View at MathSciNet
  12. D. A. Beyer and R. G. Ogier, “Tabu learning: a neural network search method for solving nonconvex optimization problems,” in Proceedings of the International Joint Conference on Neural Networks (IJCNN '91), pp. 953–961, Singapore, July 1991.
  13. K. Gopalsamy and I. Leung, “Delay induced periodicity in a neural netlet of excitation and inhibition,” Physica D, vol. 89, no. 3-4, pp. 395–426, 1996. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet
  14. L. Olien and J. Bélair, “Bifurcations, stability, and monotonicity properties of a delayed neural network model,” Physica D, vol. 102, no. 3-4, pp. 349–363, 1997. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet
  15. J. Wei and S. Ruan, “Stability and bifurcation in a neural network model with two delays,” Physica D, vol. 130, no. 3-4, pp. 255–272, 1999. View at Publisher · View at Google Scholar · View at Zentralblatt MATH · View at MathSciNet
  16. K. L. Cooke and Z. Grossman, “Discrete delay, distributed delay and stability switches,” Journal of Mathematical Analysis and Applications, vol. 86, no. 2, pp. 592–627, 1982. View at Publisher · View at Google Scholar · View at Zentralblatt MATH
  17. S. Ruan, “Absolute stability, conditional stability and bifurcation in Kolmogorov-type predator-prey systems with discrete delays,” Quarterly of Applied Mathematics, vol. 59, no. 1, pp. 159–173, 2001. View at Google Scholar · View at Zentralblatt MATH