Abstract:Attractor neural network models have been extensively studied in maintaining brain working memory state, spatial representation, and other brain processing mechanisms. The coexistence of multiple attractors in nonlinear neural network is particularly worth studying. The attractor is a stable state in the dynamic system, and all adjacent states will eventually be pulled to the attractor over time. Attractors can be regarded as the expression of data in low-dimensional state space, and different kinds of attractors can be used to process different information. The coexistence of multiple attractors means that a neural network can store multiple patterns, which means that different memory mechanisms appear at the same time. Investigating the coexistence and competition of attractors within Lotka–Volterra(LV) recurrent neural networks presents a pivotal avenue for understanding the dynamical underpinnings of complex systems, particularly in the realm of computational neuroscience and ecological modeling. This paper studies the coexistence of different types of attractors in asymmetric Lotka–Volterra recurrent neural networks. Through the weight matrix block and matrix decomposition theory, we obtain the explicit expressions of continuous attractors and discrete attractors. Since neurons have multiple grouping methods, each weight matrix block satisfying certain conditions corresponds to different attractors. Different from traditional symmetric networks, the eigenvalues of the asymmetric weight matrices may be complex, which may correspond to ring attactors. Our research shows that attractors can coexist under certain conditions, depending on the weight matrix and external input. Moreover, through simulation experiments, we show that external input will cause the competition of attractors. Finally, we provide several simulations of attractor coexistence to illustrate our theory.

Activity Invariant Sets And Stable Attractors Of Lotka-Volterra Recurrent Neural Networks

Activity Invariant Sets and Exponentially Stable Attractors of Linear Threshold Discrete-Time Recurrent Neural Networks

Attractors Coexistence and Representation Research of Lotka–Volterra Recurrent Neural Networks by Matrix Analysis Theory

Continuous Attractors of Lotka-Volterra Recurrent Neural Networks with Infinite Neurons

Nontrivial Global Attractors in 2-D Multistable Attractor Neural Networks

Continuous Attractors of Lotka-Volterra Recurrent Neural Networks

Dynamic stability conditions for Lotka-Volterra recurrent neural networks with delays.

Dynamical Properties of Continuous Attractor Neural Network with Background Tuning

Continuous Attractors of Discrete-Time Recurrent Neural Networks

Positive Invariant and Global Exponential Attractive Sets of Neural Networks with Time-Varying Delays.

Continuous Attractors and Population Decoding Multiple-Peaked Activity

Global Exponential Stability In Lagrange Sense Of Continuous-Time Recurrent Neural Networks

Multiperiodicity And Attractivity Of Delayed Recurrent Neural Networks With Unsaturating Piecewise Linear Transfer Functions

An estimation of the domain of attraction for recurrent neural networks with time-varying delays

Analysis of Continuous Attractors for 2-D Linear Threshold Neural Networks

Continuous Attractors Of A Class Of Recurrent Neural Networks Without Lateral Inhibition

Coexistence of Multiple Continuous Attractors for Lower-Ordered Neural Networks

Global convergence of Lotka-Volterra recurrent neural networks with delays

Interpreting recurrent neural networks behaviour via excitable network attractors

Representations of Continuous Attractors of Recurrent Neural Networks

Multistability and Attraction Basins of Discrete-Time Neural Networks with Nonmonotonic Piecewise Linear Activation Functions.