Computational geometry for modeling neural populations: From visualization to simulation.

Area of Science:

• Computational neuroscience
• Neural population dynamics
• Mathematical modeling of neural systems

Background:

• Mesoscopic brain descriptions are crucial for understanding neural activity.
• Existing rate-based models and single-variable population methods have limitations.
• Advanced methods are needed to capture complex neural population dynamics.

Purpose of the Study:

• To present a new method for simulating neural populations using 2D point spiking neuron models.
• To define population state via a density function over neural state space, avoiding diffusion approximations or 1D reduction.
• To enable flexible simulation of novel neural models and study noise effects in 2D systems.

Main Methods:

• Developed a simulation method for 2D point spiking neuron models.
• Represented neural population state using a density function over the neural state space.
• Implemented a modular approach allowing independent investigation of neural dynamics and stochastic processes without code recompilation.

Main Results:

• Successfully simulated neural populations using 2D models without diffusion approximation or state space reduction.
• Demonstrated the method's suitability for investigating noise in 2D systems, including regimes of large jumps.
• Showcased the ability to study novel neural models by reading state space grids.

Conclusions:

• The proposed method provides a powerful and flexible tool for studying neural population dynamics, especially in 2D systems.
• This approach offers a more comprehensive understanding of neural activity compared to 1D marginal analyses.
• Investigating noise in 2D systems complements existing 1D studies and advances computational neuroscience.