Symbol manipulation and rule learning in spiking neuronal networks

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Human language and thought exhibit productivity, systematicity, and compositionality, suggesting a need for a physical symbol system (PSS) in the brain.
• Temporal coding in neurons offers a new avenue for implementing PSS.
• Learning classifier systems provide a computational framework for symbol manipulation and rule learning.

Purpose of the Study:

• To propose and describe a novel neuronal implementation of a physical symbol system (PSS).
• To integrate algorithmic and implementation-level descriptions of a neuronal system supporting systematic and compositional behaviors.
• To compare the proposed model with existing neuronal implementations of symbolic representations.

Main Methods:

• Investigating temporal coding for neuronal PSS implementation.
• Utilizing learning classifier systems to model symbol re-write rules and natural selection.
• Implementing the core operations of learning classifier systems via spike-time dependent plasticity (STDP) based supervised learning.

Main Results:

• Demonstrated a plausible algorithmic basis for systematic and compositional behaviors using learning classifier systems.
• Showcased how STDP-based supervised learning can implement the replication with variation of symbol re-write rules.
• Presented a unified model integrating algorithmic and neuronal implementation levels for a symbolic system.

Conclusions:

• The proposed model offers a novel neuronal mechanism for a physical symbol system (PSS) in the brain.
• This approach integrates computational principles with neurobiological mechanisms to explain cognitive functions like systematicity and compositionality.
• The study provides a framework for understanding how neuronal networks can support symbolic thought and behavior.