Long-term memory stabilized by noise-induced rehearsal

Area of Science:

• Neuroscience
• Computational Neuroscience
• Neural Network Modeling

Background:

• Cortical networks store memories long-term, yet synapses have short lifespans.
• The mechanism for maintaining memory stability with transient synaptic strengths remains unclear.

Purpose of the Study:

• Investigate if neural networks can store enduring memories using unstable synapses.
• Examine the impact of ongoing spike-timing-dependent plasticity (STDP) on memory pattern stability.

Main Methods:

• Utilized an attractor neural network model.
• Simulated the effects of various STDP rules on synaptic plasticity.
• Analyzed temporal correlations in neural activity resulting from network interactions and noise.

Main Results:

• Identified specific STDP rules that stabilize all stored memory patterns.
• Demonstrated that neural noise, processed by recurrent connections, imprints memory patterns.
• Showed that STDP reinforces all stored patterns via these temporal correlations, including unvisited ones.

Conclusions:

• Ongoing STDP can stabilize memory patterns in neural networks with short-lived synapses.
• Irregular neural activity, driven by STDP and network correlations, is crucial for maintaining stable neural connections and memory consolidation.
• Findings support the hypothesis that irregular spiking is a key feature for stable cortical networks.