Discrete Synaptic Events Induce Global Oscillations in Balanced Neural Networks

Area of Science:

• Computational Neuroscience
• Theoretical Neuroscience
• Network Dynamics

Background:

• Neural dynamics are driven by discrete synaptic events.
• Current models often use diffusion approximation, treating synaptic inputs as Gaussian noise.
• This simplification may not capture all emergent network behaviors.

Purpose of the Study:

• To develop a mean-field formalism that includes synaptic shot noise.
• To analyze neural dynamics in sparse balanced networks.
• To compare the predictions of the new model with the diffusion approximation.

Main Methods:

• Derivation of a mean-field formalism incorporating synaptic shot noise.
• Analysis of sparse balanced neural networks.
• Investigation of network transitions under varying excitatory drive and inhibitory feedback.

Main Results:

• The developed formalism accurately predicts global oscillations emerging via continuous or hysteretic transitions.
• These oscillations are observed under low excitatory drive and high inhibitory feedback.
• The diffusion approximation fails to predict these emergent oscillations.
• At low in-degrees, oscillation nature shifts from drift-driven to cluster activation.

Conclusions:

• Synaptic shot noise is crucial for accurately modeling emergent global oscillations in neural networks.
• The derived mean-field approach provides a more accurate description than the diffusion approximation.
• Network behavior, including oscillation characteristics, is sensitive to input statistics and network topology.