Abstract:It is widely believed that memory storage depends on activity-dependent synaptic modifications. Classical studies of learning and memory in neural networks describe synaptic efficacy either as continuous or discrete. However, recent results suggest an intermediate scenario in which synaptic efficacy can be described by a continuous variable, but whose distribution is peaked around a small set of discrete values. Motivated by these results, we explored a model in which each synapse is described by a continuous variable that evolves in a potential with multiple minima. External inputs to the network can switch synapses from one potential well to another. Our analytical and numerical results show that this model can interpolate between models with discrete synapses which correspond to the deep potential limit, and models in which synapses evolve in a single quadratic potential. We find that the storage capacity of the network with double well synapses exhibits a power law dependence on the network size, rather than the logarithmic dependence observed in models with single well synapses. In addition, synapses with deeper potential wells lead to more robust information storage in the presence of noise. When memories are sparsely encoded, the scaling of the capacity with network size is similar to previously studied network models in the sparse coding limit. A long-lasting question in neuroscience is whether synaptic efficacies should be described as continuous variable or discrete variables. Recent experiments indicate that it is a combination of both: synaptic efficacy changes continuously, but its distribution peaks at several discrete values. In this study, we introduce a synapse model described by a double well potential, and investigate the memory properties of networks of neurons connected with such synapses. Our results show in networks with a bimodal weight distribution, the storage capacity depends on network size as a power law. In addition, we demonstrate that networks with such synapses store information more robustly in the presence of noise, compared to networks with synapses with a single well potential.

Large Memory Capacity in Chaotic Artificial Neural Networks: A View of the Anti-Integrable Limit.

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

Controlling Chaos In A Neural Network Based On The Phase Space Constraint

A Type of Delay Feedback Control of Chaotic Dynamics in A Chaotic Neural Network

Sinusoidal Modulation Control Method in a Chaotic Neural Network

Chaos in A Neuron Model with Adaptive Feedback Synapse: Electronic Circuit Design

Associative memory with a controlled chaotic neural network.

A Chaotic Associative Memory

Chaotic and Hyperchaotic Attractors in Time-Delayed Neural Networks

2008 Special Issue: Threshold control of chaotic neural network

Memorizing morph patterns in small-world neuronal network

Attractor neural networks with double well synapses

Lyapunov Exponent and Chaotic Area Distribution of a Chaotic Neural Network

Dynamics Analysis and Analog Associative Memory of Networks with LT Neurons.

Review on chaotic dynamics of memristive neuron and neural network

Efficient Spiking Neural Networks with Sparse Selective Activation for Continual Learning

Studying a Chaotic Spiking Neural Model

Expressivity of Neural Networks via Chaotic Itineraries beyond Sharkovsky's Theorem

A new class of chaotic attractors using different activation functions in neuron with multi dendrites

A note on maximum independent sets in rectangle intersection graphs

Chaotic phenomenon in both an integrate-and-fire circuit with periodic pulse-train input and a discrete neural network model