Stimulus number, duration and intensity encoding in randomly connected attractor networks with synaptic depression.

Area of Science:

• Computational neuroscience
• Neural network dynamics
• Synaptic plasticity

Background:

• Recurrent neural networks with excitatory neurons exhibit multiple attractor states.
• Synaptic depression can destabilize these attractors with increased input, leading to network activity changes.

Purpose of the Study:

• To investigate how depressing synapses in recurrent neural networks influence attractor state dynamics.
• To determine if these networks can encode stimulus properties like number, duration, and amplitude.

Main Methods:

• Simulated randomly connected recurrent networks of excitatory neurons with depressing synapses.
• Analyzed network activity patterns in response to transient stimuli of varying number, duration, and amplitude.

Main Results:

• Depressing synapses induce itinerancy, causing networks to traverse sequences of attractor states.
• Networks encode stimulus number via distributed neural activity with non-monotonic tuning curves.
• Stimulus duration and amplitude are encoded through distinct distributed representations, differentiating them from stimulus number.

Conclusions:

• Short-term synaptic depression is crucial for networks to dynamically encode and distinguish stimulus properties.
• This mechanism allows neural networks to process complex temporal and intensity information beyond simple integration.