Learning the structure of correlated synaptic subgroups using stable and competitive spike-timing-dependent...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Synaptic plasticity is crucial for continuous learning of neural input structures.
• This requires a balance between competitive and stable synaptic modifications.
• Existing models often struggle to achieve both properties simultaneously.

Purpose of the Study:

• To identify a class of spike-timing-dependent plasticity (STDP) models with both competitive and stable properties.
• To investigate synaptic structure formation under correlated input subgroups.
• To determine the influence of learning rates on synaptic structure stability and neuronal response properties.

Main Methods:

• Analysis of a wide class of spike-timing-dependent plasticity (STDP) models.
• Mathematical modeling of synaptic structure formation.
• Simulation of neuronal responses under varying learning rates and input correlations.

Main Results:

• A specific class of STDP models demonstrates both stability and competitiveness for correlated inputs.
• Small learning rates yield multiple possible synaptic structures; large rates prevent stabilization.
• Intermediate learning rates typically form a unique, stable synaptic structure, enabling selective neuronal responses.
• Selectivity robustness depends on the ratio of subgroup correlation strength to the number of subgroups.
• The balance between potentiation and depression dictates the fraction of potentiated subgroups.

Conclusions:

• A specific class of STDP models can achieve stable and competitive synaptic plasticity for structured inputs.
• Optimal learning rates are essential for forming unique and stable synaptic structures, leading to selective neuronal function.
• The interplay between correlation strength, subgroup number, and potentiation/depression balance governs learning outcomes.