Maximum memory capacity on neural networks with short-term synaptic depression and facilitation

Area of Science:

• Computational Neuroscience
• Artificial Neural Networks
• Synaptic Plasticity

Background:

• Neural networks rely on synaptic processes for memory storage.
• Activity-dependent synaptic processes like facilitation and depression dynamically alter synaptic strength.
• Understanding these dynamics is crucial for optimizing neural network memory capacity.

Purpose of the Study:

• To investigate the impact of competing synaptic processes (facilitation and depression) on neural network memory storage capacity.
• To identify optimal synaptic parameters for maximizing memory storage.
• To compare the storage capacity of dynamic synapses with static synapses.

Main Methods:

• Analytical studies of synaptic processes.
• Monte Carlo simulations of neural network models.
• Analysis of synaptic parameters including neurotransmitter release probability and facilitation time constant.

Main Results:

• Synaptic depression significantly reduces memory storage capacity for static patterns.
• Synaptic facilitation enhances storage capacity in various contexts.
• Optimal synaptic parameter values were found, achieving maximal storage capacity comparable to static synapses.

Conclusions:

• Depressing synapses with a degree of facilitation can restore the storage properties of static synapses.
• This approach preserves the nonlinear characteristics of dynamic synapses, beneficial for information processing.
• The findings suggest a mechanism for robust information coding in neural networks.