Modeling and contractivity of neural-synaptic networks with Hebbian learning
Motivated by some biological considerations, neural synaptic systems that combine Hopfield neural networks (HNN) and firing -- rate neural networks (FRNN) for the neural dynamics with two different Hebbian learning (HL) rules for the synaptic dynamics, are studied. A review of some basics as: contraction, composite norms, Metzler matrix, Out-incidence and in-incidence matrices, is done in the second section of the paper. The third section is devoted to the presentation of the models. The dynamics of Hopfield neural networks, firing-rate neural networks as well as the Hebbian learning rule (HL), the anti-Hebbian learning rule and the Oja's like learning rule are defined. Using these neural and synaptic dynamics three coupled neural synaptic models are considered. The first one, Hopfield-Hebbian model, combines the HNN dynamics with the HL learning rule, under some initial neural and synaptic conditions. The second one, the firing-rate-Hebbian model, combines the FRNN dynamics with the HL learning rule, under some initial neural and synaptic conditions. The third one, the Hopfield-Oja model, is the combination of the HNN dynamics with the Oja's like synaptic plasticity rule, still under some initial neural and synaptic conditions. Assumptions on bounds for activation functions and external stimuli are introduced. Further, these three models will be written in a low-dimensional formulation. In the fourth section one investigates the boundedness of solutions to the models, the contractivity of the models and whether the models satisfy the Dale's principle. The theoretical results in this article are validated by an example for the Hopfield-Hebbian model. Conclusions and future work are presented in the last section of the article.\N\NAn interesting and comprehensive article.
- Stabilization of Hebbian neural nets by inhibitory learning
- A Mathematical Analysis of the Effects of Hebbian Learning Rules on the Dynamics and Structure of Discrete-Time Random Recurrent Neural Networks
- Constraints on learning in dynamic synapses
- Neural networks with fast time-variation of synapses
- Dynamics of Hopfield neural networks with dynamic synapses
- A Contraction Approach to the Hierarchical Analysis and Design of Networked Systems
- A Differential Lyapunov Framework for Contraction Analysis
- A simplified neuron model as a principal component analyzer
- Contraction Analysis of Time-Delayed Communications and Group Cooperation
- Deep learning
- Dynamic properties of neural networks with adapting synapses
- Hebbian Learning of Recurrent Connections: A Geometrical Perspective
- Hierarchical Selective Recruitment in Linear-Threshold Brain Networks—Part I: Single-Layer Dynamics and Selective Inhibition
- scientific article; zbMATH DE number 3100403 (Why is no real title available?)
- Learning representations by back-propagating errors
- Mathematical equivalence of two common forms of firing rate models of neural networks
- Mathematical formulations of Hebbian learning
- Matrix measures, stability and contraction theory for dynamical systems on time scales
- Neurons with graded response have collective computational properties like those of two-state neurons
- New conditions for global stability of neural networks with application to linear and quadratic programming problems
- Non-Euclidean Contraction Theory for Robust Nonlinear Stability
- Nonlinear systems.
- On contraction analysis for non-linear systems
- On Logarithmic Norms
- Transformations by diagonal matrices in a normed space
- Unsupervised learning by competing hidden units
This page was built for publication: Modeling and contractivity of neural-synaptic networks with Hebbian learning
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6550233)
