The Impact of Sparse Coding on Memory Lifetimes in Simple and Complex Models of Synaptic Plasticity

Area of Science:

• Computational Neuroscience
• Memory Systems

Background:

• Associative memory models with simple synapses suffer from rapid memory forgetting.
• Sparse coding and complex synaptic plasticity (metaplasticity) are proposed to enhance memory duration.

Purpose of the Study:

• To investigate memory lifetimes in simple and complex synaptic plasticity models under sparse coding.
• To compare filter-based, cascade, and serial synapse models for memory retention.

Main Methods:

• Examined memory lifetimes at single-neuron and population levels using signal-to-noise ratio (SNR) and first passage time (FPT) methods.
• Analyzed the impact of synaptic complexity, sparseness, and spontaneous activity on memory duration.

Main Results:

• Optimal single-neuron memory lifetime decreases with increasing synaptic complexity, irrespective of spontaneous activity.
• Population-level memory lifetimes increase with synaptic complexity in filter and serial models, critically dependent on spontaneous activity.
• SNR validity is challenged by sparse coding, particularly for complex synaptic plasticity models.

Conclusions:

• Synaptic complexity presents a trade-off between single-neuron and population-level memory performance.
• Spontaneous activity is crucial for enhancing memory lifetimes at the population level with complex synapses.
• The validity of SNR for memory lifetime estimation requires careful consideration, especially in highly sparse regimes.