Multiple topological representation self-organized by spike-timing-dependent synaptic learning rule.

Area of Science:

• Computational neuroscience
• Neuroscience
• Artificial intelligence

Background:

• Neurons in the inferior temporal (IT) cortex are responsible for position- and scale-free shape recognition.
• Familiar shapes occupy a restricted, clustered region within the vast space of all possible shapes, exhibiting both continuity and discreteness.

Purpose of the Study:

• To demonstrate a spike-based model capable of acquiring multiple representations.
• To investigate how topological maps based on spike-timing-dependent synaptic plasticity (STDP) can model complex shape representations.

Main Methods:

• Utilized a spike-based computational model incorporating STDP.
• Simulated inputs on multiple rings, representing a simplified mixed structure of shape space.
• Analyzed the resulting neural activity patterns and synaptic plasticity.

Main Results:

• The model successfully acquired multiple representations, with position on rings represented by active neuron centers and ring differences by detailed activity patterns.
• Neurons in proximity showed high activity for inputs from other rings, reflecting a distributed representation.
• The model demonstrated continuous shifts in active neuron regions corresponding to continuous input changes.

Conclusions:

• The developed spike-based model effectively captures the mixed structure of shape representations observed in the IT cortex.
• The findings support the idea that STDP can underlie the emergence of complex, continuous, and discrete representations in neural systems.
• The model provides a framework for understanding how the brain processes and represents complex visual information.