From Dyson to Hopfield: Processing on hierarchical networks

Abstract

We consider statistical-mechanical models for spin systems built on hierarchical structures, which provide a simple example of non-mean-field framework. We show that the coupling decay with spin distance can give rise to peculiar features and phase diagrams much richer that their mean-field counterpart. In particular, we consider the Dyson model, mimicking ferromagnetism in lattices, and we prove the existence of a number of meta-stabilities, beyond the ordered state, which get stable in the thermodynamic limit. Such a feature is retained when the hierarchical structure is coupled with the Hebb rule for learning, hence mimicking the modular architecture of neurons, and gives rise to an associative network able to perform both as a serial processor as well as a parallel processor, depending crucially on the external stimuli and on the rate of interaction decay with distance; however, those emergent multitasking features reduce the network capacity with respect to the mean-field counterpart. The analysis is accomplished through statistical mechanics, graph theory, signal-to-noise technique and numerical simulations in full consistency. Our results shed light on the biological complexity shown by real networks, and suggest future directions for understanding more realistic models.

References in corpus (7)

• Griffiths phases and the stretching of criticality in brain networks
• Scaling theory of transport in complex networks
• Multitasking associative networks
• Extensive load in multitasking associative networks
• The Hierarchical Random Energy Model
• Meta-stable states in the hierarchical Dyson model drive parallel processing in the hierarchical Hopfield network
• Free-energy bounds for hierarchical spin models

Cited by in corpus (29)

• Regulation of Irregular Neuronal Firing by Autaptic Transmission
• Information Perspective to Probabilistic Modeling: Boltzmann Machines versus Born Machines
• Hierarchical organization of functional connectivity in the mouse brain: a complex network approach
• Non-Convex Multi-species Hopfield models
• Neural Networks retrieving Boolean patterns in a sea of Gaussian ones
• Deterministic hierarchical networks
• Hierarchical neural networks perform both serial and parallel processing
• The ambiguity of nestedness under soft and hard constraints
• Avalanche precursors of failure in hierarchical fuse networks
• Topological dimension tunes activity patterns in hierarchical modular network models
• Integrable extended van der Waals model
• First-passage phenomena in hierarchical networks
• Hopfield model with planted patterns: a teacher-student self-supervised learning model
• Parallel Learning by Multitasking Neural Networks
• Topological properties of hierarchical networks
• Real-Space Renormalization Group for Spectral Properties of Hierarchical Networks
• Fast Reversible Learning based on Neurons functioning as Anisotropic Multiplex Hubs
• Dense Hopfield Networks in the Teacher-Student Setting
• Graph theoretical approaches for the characterization of damage in hierarchical materials
• The effect of priors on Learning with Restricted Boltzmann Machines
• Modeling Structured Data Learning with Restricted Boltzmann Machines in the Teacher-Student Setting
• The Relativistic Hopfield network: rigorous results
• Regular graphs maximize the variability of random neural networks
• The relativistic Hopfield model with correlated patterns
• Dynamics of one-dimensional spin models under the line-graph operator
• Recognition Capabilities of a Hopfield Model with Auxiliary Hidden Neurons
• Evaluation of Spectral Zeta-Functions with the Renormalization Group
• Pattern recognition in Deep Boltzmann machines
• Response time of lateral predictive coding and benefits of modular structures