Abstract
In this paper, the fixed-time bipartite synchronization problem for coupled delayed neural networks with signed graphs is discussed. Different from traditional neural networks, the interactions between nodes of delayed neural networks can be either collaborative or antagonistic. Furthermore, compared with the initial-condition based finite-time synchronization, the settling time is bounded by a constant within fixed-time regardless of the initial condition. It is worth noting that the fixed-time stable network for bipartite synchronization in this paper achieves faster convergence than most existing publications. By applying the constructing comparison system method, Lyapunov stability theory, and inequality techniques, some sufficient criteria for fixed-time bipartite synchronization are obtained. Finally, two numerical examples are granted to display the performance of the obtained results.
Introduction
In the field of network science, synchronization is a fundamental phenomenon where individual nodes in a system eventually converge to a common behavior. While traditional models often assume purely collaborative interactions, real-world systems—ranging from social networks with conflicting interests to biological systems with inhibitory neurons—frequently involve antagonistic relationships. This complexity necessitates the study of bipartite synchronization, where a network splits into two distinct groups: nodes within a group synchronize with each other, while nodes in opposing groups mirror each other with opposite signs.
This research addresses three critical challenges in the synchronization of neural networks:
- Signed Graphs: Accounting for both positive (collaborative) and negative (antagonistic) links.
- Time Delays: Managing the inherent delays in signal transmission that often destabilize neural systems.
- Convergence Speed: Moving beyond "finite-time" synchronization, which depends on initial states, to a "fast" fixed-time approach that guarantees synchronization within a specific deadline regardless of the starting conditions.
Methodology
The study employs a rigorous mathematical framework to design and validate synchronization criteria. The core of the approach lies in Algebraic Graph Theory, specifically focusing on structurally balanced signed graphs. These graphs are essential for bipartite synchronization as they allow the network to be partitioned into two opposing sets without logical paradoxes (e.g., "cycles of enemies").
To ensure stability and estimate convergence, the researchers utilized Lyapunov Stability Theory and the Comparison System Method. By comparing the complex neural network to a simpler differential equation, they could derive the conditions under which synchronization error reaches zero.
A pivotal contribution of this work is the Novel Controller Design. The researchers developed a feedback controller ($u_i(t)$) incorporating non-linear terms, such as sign functions and power laws. By utilizing a specific mathematical lemma (Lemma 3) that provides a tighter bound on settling time than standard methods, the authors achieved faster convergence than previously documented in the literature.
Numerical Validation
The theoretical results were validated through two distinct numerical simulations involving 9-node networks:
- Example 1: A network coupled by a chaotic system with a coupling strength ($\sigma$) of 0.2. The simulation demonstrated that the network reached bipartite synchronization in approximately 1.2176 seconds, well within the predicted theoretical bounds.
- Example 2: This setup utilized a different activation function (a saturation function) to test the robustness of the criteria. In this instance, the network achieved synchronization even faster, in roughly 0.5853 seconds.
In both cases, the results confirmed that the proposed controller effectively forces nodes into synchronized opposition within a guaranteed timeframe, independent of their initial states.
Conclusion and Future Directions
This paper provides a robust mathematical foundation for achieving fast fixed-time bipartite synchronization in delayed neural networks. While the results are significant, the authors identify several avenues for future research. Currently, the theory relies on the network being structurally balanced; however, real-world networks are often "unbalanced," leading to more complex frustrated states. Additionally, future work will aim to scale these controllers for massive, high-dimensional networks and address time-varying delays and impulsive control effects, where adjustments are made in short bursts rather than continuous signals.
Summary updated