Integration of Continuous-Time Dynamics in a Spiking Neural Network Simulator

Area of Science:

• Computational Neuroscience
• Computational Neuroscience and Neural Networks

Background:

• Contemporary neural network modeling utilizes two main approaches: biologically grounded spiking neuron models and functionally inspired rate-based models.
• Integrating these distinct modeling paradigms for multi-scale simulations presents significant technical challenges.

Purpose of the Study:

• To present a unified simulation framework enabling the combination of spiking neuron models and rate-based models.
• To facilitate quantitative validation of mean-field approaches using spiking network simulations.
• To enhance simulation reliability through consistent code and model specifications for both model classes.

Main Methods:

• Developed a unified simulation framework supporting multi-scale modeling by integrating spiking and rate-based neural network models.
• Implemented instantaneous and delayed interactions for rate-based models within a spiking network simulator, drawing parallels to gap junction inclusion.
• Introduced an iterative waveform-relaxation technique to optimize large-scale rate-based model simulations.

Main Results:

• The framework successfully integrates rate-based and spiking neuron models, allowing for joint simulations.
• Quantitative validation of mean-field approaches against spiking network simulations is enabled.
• Demonstrated broad applicability across various network models, from random networks to neural-field models.

Conclusions:

• The unified framework provides a prerequisite for seamless interaction between rate-based and spiking neural network models.
• This approach enhances the reliability and flexibility of multi-scale neural network simulations.
• The study paves the way for more comprehensive and accurate computational neuroscience research.