Abstract:We extend a recently introduced class of exactly solvable models for recurrent neural networks with competition between 1D nearest neighbour and infinite range information processing. We increase the potential for further frustration and competition in these models, as well as their biological relevance, by adding next-nearest neighbour couplings, and we allow for modulation of the attractors so that we can interpolate continuously between situations with different numbers of stored patterns. Our models are solved by combining mean field and random field techniques. They exhibit increasingly complex phase diagrams with novel phases, separated by multiple first- and second order transitions (dynamical and thermodynamic ones), and, upon modulating the attractor strengths, non-trivial scenarios of phase diagram deformation. Our predictions are in excellent agreement with numerical simulations.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is related to the modulation and proliferation of attractors in neural networks. Specifically, the author has extended a class of recently proposed exactly solvable recurrent neural network models, which include one - dimensional nearest - neighbor and infinite - range information - processing competition. By adding next - nearest - neighbor coupling and allowing the modulation of attractors to achieve continuous interpolation between different numbers of stored patterns, the paper aims to: 1. **Enhance the biological relevance of the model**: By introducing next - nearest - neighbor coupling, the model can better simulate the local and long - range information - processing mechanisms in real biological neural networks, thereby improving the biological realism of the model. 2. **Explore new features of the phase diagram**: By introducing new interaction parameters, study how these parameters affect the phase diagram of the model, especially what new phases will be produced when these parameters lead to further competition and frustration in the network. 3. **Understand the impact of changes in the number of stored patterns**: By modulating the embedding strength of stored patterns, study how the phase diagram will change significantly when the number of stored patterns increases from 1 to 2. This helps to understand the disruptive effect of non - condensed patterns on the recall of condensed patterns. ### Main contributions - **Model extension**: The paper proposes two orthogonal extension methods, one is to add next - nearest - neighbor coupling, and the other is to modulate the embedding strength of stored patterns. - **Rich phase diagram**: Through these extensions, the model shows a more complex phase diagram, including multiple types of phase transitions (first - order and second - order phase transitions) and different numbers of steady - state solutions. - **Consistency between theory and numerical simulation**: The theoretical predictions in the paper are highly consistent with the numerical simulation results, verifying the effectiveness of the model. ### Technical details - **Model definition**: The model is based on the Ising spin system and evolves according to the Glauber dynamics rules. The interaction matrix \(J_{ij}\) contains infinite - range and short - range interactions. - **Free - energy calculation**: Derive the expression of free energy through the random - field technique and solve it using the saddle - point method. - **Phase - diagram analysis**: Determine different phase regions and their boundaries by numerically calculating the local minima of the free - energy surface. ### Conclusion Through the extension and in - depth analysis of the model, the paper reveals the complex behavior of attractor modulation and proliferation in neural networks, especially the phase - transition phenomena under different numbers of stored patterns and interaction parameters. These findings not only enrich the theoretical model but also provide a new perspective for understanding the dynamic behavior of real biological neural networks.

Attractor Modulation and Proliferation in 1+$\infty$ Dimensional Neural Networks

Multiperiodicity of Periodically Oscillated Discrete-Time Neural Networks with Transient Excitatory Self-Connections and Sigmoidal Nonlinearities

Nontrivial Global Attractors in 2-D Multistable Attractor Neural Networks

Models of Innate Neural Attractors and Their Applications for Neural Information Processing

Coexisting Chaotic Attractors in A Single Neuron Model with Adapting Feedback Synapse

Attractor-based models for sequences and pattern generation in neural circuits

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

Feed-Forward Chains of Recurrent Attractor Neural Networks Near Saturation

Attractor Stability in Finite Asynchronous Biological System Models

Slow and Weak Attractor Computation Embedded in Fast and Strong E-I Balanced Neural Dynamics.

Spontaneous Motion on Two-dimensional Continuous Attractors

Coexisting firing patterns and attractor selection in memristive synapse coupled heterogeneous neurons

Back to the Continuous Attractor

Coexistence of continuous attractors with different dimensions for neural networks

Spiral Attractors in a Reduced Mean-Field Model of Neuron-Glial Interaction

A hybrid model for large neural network description

Simultaneous embedding of multiple attractor manifolds in a recurrent neural network using constrained gradient optimization

Revealing and reshaping attractor dynamics in large networks of cortical neurons

Neural information processing with feedback modulations.

Infinite Neural Network Quantum States: Entanglement and Training Dynamics

Non-equilibrium statistical mechanics of continuous attractors