Computer simulation of inhibition-dependent binding in a neural network.

Area of Science:

• Computational Neuroscience
• Artificial Neural Networks
• Systems Neuroscience

Background:

• Neural networks exhibit complex dynamics, with inhibition playing a crucial role in network function.
• Binding mechanisms in neural systems are essential for information processing and cognitive functions.
• Previous models have explored temporal coherence in synaptic inputs for neural triggering.

Purpose of the Study:

• To model reverberating dynamics in a neural network using binding neurons.
• To investigate the role of inhibition as a binding controller within the network.
• To explore learning mechanisms for controlling network activity patterns.

Main Methods:

• A neural network model composed of binding neurons was simulated on a PC.
• The binding neuron model utilizes temporal coherence of synaptic inputs for triggering.
• Two learning mechanisms were implemented: adjusting synaptic strength and propagation delays.
• External patterns were used to train the network to support distinct activity dynamics.

Main Results:

• The network demonstrated the ability to support both disconnected and bound patterns of activity.
• High inhibition levels led to disconnected activity patterns.
• Low inhibition levels resulted in bound activity patterns.
• The system successfully learned to switch between these patterns based on inhibition levels.

Conclusions:

• Inhibition acts as a critical controller for binding phenomena in this neural network model.
• The degree of slow inhibition directly influences the network's ability to form bound activity patterns.
• This model provides insights into how inhibition shapes network dynamics and information integration.