A neural model of perceptual-motor alignment

Area of Science:

• Neuroscience
• Computational Neuroscience
• Sensorimotor Systems

Background:

• Sensorimotor systems constantly adjust to spatial discrepancies caused by factors like growth and optical distortion.
• Adaptation involves establishing coherent perceptual-motor alignments for a stable perception of the world.
• Previous studies show adaptation links entire modalities, not just isolated instances.

Purpose of the Study:

• To investigate how biological neural networks computationally solve the problem of inducing constrained relations between continuous stimulus and response dimensions.
• To explore the mechanisms underlying perceptual-motor adaptations in sensorimotor systems.
• To model the interpolation and extrapolation capacities of neural networks in sensorimotor adaptation.

Main Methods:

• Experimental manipulations exposing participants to localized spatial discrepancies.
• Developing a computational model based on linear collective computation and least-square (LS) error learning.
• Utilizing populations of frequency-coded neurons where discharge varies monotonically with a parameter.

Main Results:

• Adaptation is characterized by the acquisition of constrained relations between entire spatial modalities.
• The proposed neural network model demonstrates inherent interpolation and extrapolation capacities.
• The model successfully accounts for the observed properties of perceptual-motor adaptations.

Conclusions:

• Neural processing using linear collective computation and LS error learning provides a framework for understanding sensorimotor adaptation.
• This model explains how the brain performs interpolation and extrapolation to maintain perceptual-motor alignment.
• The findings offer insights into the computational principles governing biological neural networks in sensorimotor adaptation.