Cortical Motion Perception Emerges from Dimensionality Reduction with Evolved Spike-Timing-Dependent Plasticity...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Processing

Background:

• The nervous system faces energy constraints, necessitating efficient information representation.
• Neurons in the medial superior temporal area (MSTd) of primates encode complex motion patterns crucial for behavior.
• Previous models suggest dimensionality reduction via non-negative matrix factorization (NMF) explains MSTd response properties.

Purpose of the Study:

• To investigate how neural circuits implement efficient coding and dimensionality reduction in MSTd.
• To propose and analyze a spiking neural network (SNN) model of MSTd.
• To explore the role of spike-timing-dependent plasticity and homeostatic synaptic scaling (STDP-H) in learning efficient representations.

Main Methods:

• Developed a spiking neural network (SNN) model for the MSTd.
• Incorporated evolved spike-timing-dependent plasticity and homeostatic synaptic scaling (STDP-H) learning rules.
• Analyzed the emergent representations and receptive fields of simulated MSTd neurons.

Main Results:

• The SNN model learned compressed and efficient representations of visual input patterns.
• These representations were similar to those generated by non-negative matrix factorization (NMF).
• The model produced MSTd-like receptive fields observed in biological recordings.

Conclusions:

• MSTd's complex neuronal responses may arise from dimensionality reduction and efficient coding.
• Spike-timing-dependent plasticity with homeostatic synaptic scaling (STDP-H) can implement functions similar to NMF.
• This SNN model provides a mechanistic explanation for efficient visual motion encoding in MSTd for self-motion perception.