Stable Neural Population Dynamics in the Regression Subspace for Continuous and Categorical Task Parameters in...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Neural population dynamics offer a computational framework for brain function, representing activity as low-dimensional trajectories.
• Conventional rate-coding analysis, focusing on single-neuron firing rates, is poorly integrated with dynamic models.
• Bridging these models is crucial for a comprehensive understanding of neural information processing.

Purpose of the Study:

• To develop a method linking rate-coding and dynamic models of neural activity.
• To analyze how task parameters influence neural population dynamics.
• To explore the geometric representation of neural modulations in response to stimuli.

Main Methods:

• Developed a state-space analysis variant within the regression subspace.
• Applied the method to macaque monkey neural population datasets with continuous and categorical task parameters.
• Integrated optimal-stimulus response analysis with dynamic modeling.

Main Results:

• Neural modulation structures were reliably captured by task parameters in the regression subspace as low-dimensional trajectory geometries.
• Prominent modulation dynamics in the lower dimension originated from optimal responses.
• Extracted geometries for both continuous and categorical task parameters formed a straight geometry, indicating unidimensional functional relevance.

Conclusions:

• The developed approach successfully bridges neural modulation in rate-coding and dynamic systems.
• This method provides a significant advantage for exploring the temporal structure of neural modulations in existing datasets.
• The findings suggest that functional relevance in neural modulation dynamics can be characterized as a unidimensional feature.