Synergies between intrinsic and synaptic plasticity mechanisms

Area of Science:

• Computational neuroscience
• Information theory
• Machine learning

Background:

• Neurons exhibit intrinsic plasticity, a mechanism distinct from synaptic plasticity.
• The BCM theory describes synaptic plasticity with a sliding threshold, but its underlying mechanism remains debated.
• Understanding neural plasticity is crucial for deciphering learning and information processing in the brain.

Purpose of the Study:

• To propose an information-theoretic model of intrinsic plasticity for a continuous activation neuron.
• To investigate the interplay between intrinsic and synaptic plasticity.
• To explore intrinsic plasticity as an alternative explanation for the BCM theory's sliding threshold.

Main Methods:

• Developed an information-theoretic model of intrinsic plasticity.
• Analyzed the interaction between intrinsic plasticity and synaptic plasticity.
• Theoretically analyzed intrinsic plasticity with Hebbian learning rules for clustered inputs.
• Conducted experiments on the nonlinear independent component analysis 'bars' problem.

Main Results:

• Demonstrated that the interaction of intrinsic and synaptic plasticity enables neurons to discover heavy-tailed input distributions.
• Showed that intrinsic plasticity can serve as an alternative mechanism for the sliding threshold in synaptic plasticity (BCM theory).
• Experimental results on the 'bars' problem validated the theoretical findings.

Conclusions:

• Intrinsic plasticity plays a significant role in neural computation, complementing synaptic plasticity.
• The proposed model offers a novel perspective on neural learning and adaptation.
• This work provides a theoretical and experimental framework for understanding complex neural plasticity mechanisms.