The predictive brain: temporal coincidence and temporal order in synaptic learning mechanisms.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Learning Theory

Background:

• Synaptic plasticity, crucial for learning, often relies on coincident pre- and postsynaptic activity, aligning with Hebbian (correlational) learning.
• Volume learning suggests synaptic plasticity may involve diffusible signals, incorporating spatial aspects into correlational rules.
• Existing models are largely insensitive to the precise temporal order of synaptic inputs, despite behavioral evidence of temporal sensitivity in animals.

Purpose of the Study:

• To propose a new class of learning rule, the predictive Hebbian learning rule, that incorporates temporal ordering of synaptic inputs.
• To explore how this predictive learning rule can operate at the level of single synapses and through neuromodulatory systems.
• To bridge the gap between correlational learning rules and observed temporal sensitivity in biological systems.

Main Methods:

• Theoretical modeling of neural network learning rules.
• Analysis of existing evidence from conditioning experiments and psychophysical studies.
• Conceptual framework for predictive Hebbian learning mechanisms.

Main Results:

• A novel predictive Hebbian learning rule is proposed, which is sensitive to the temporal sequence of synaptic inputs.
• This rule can be implemented at individual synaptic connections.
• The rule can also function via diffuse neuromodulatory systems, impacting broader neural circuits.

Conclusions:

• Predictive Hebbian learning offers a mechanism for neural networks to process temporally ordered information.
• This framework potentially explains the sensitivity to temporal order observed in animal behavior and perception.
• The proposed rule extends Hebbian principles to account for temporal dynamics in synaptic plasticity and learning.