[Information characteristics of neuronal and synaptic plasticity]

Area of Science:

• Neuroscience
• Computational Neuroscience
• Information Theory

Context:

• Neuronal networks (NN) with plastic elements are crucial for understanding brain function and developing artificial intelligence.
• Assessing informational losses is key to optimizing NN performance and capacity.
• The study investigates how different types of synaptic plasticity affect information processing in NNs.

Purpose:

• To estimate informational losses in neuronal networks (NNs) with plastic elements.
• To analyze the impact of decoding strategies (complicated vs. simple) on information loss.
• To compare information capacity of NNs with gradual, Olbus-type, and Hebbian plasticity.

Summary:

• Informational losses arise from the transition from complex to simple decoding and structural uncertainties in plastic NNs.
• Gradual plasticity leads to substantial information loss, diminishing the NN's advantage over binary networks.
• Olbus-type plasticity shows no significant informational losses, while Hebbian plasticity exhibits high losses dependent on network parameters.

Impact:

• Highlights the limitations of gradual plasticity in maintaining information capacity in NNs.
• Suggests that Olbus-type plasticity is more efficient for information preservation in neural computations.
• Provides insights into the parameter-dependent nature of information loss in Hebbian-type plastic synapses.