Collective behavior of large-scale neural networks with GPU acceleration

Area of Science:

• Computational neuroscience
• Neuroscience modeling

Background:

• Mammalian cortex anatomy inspires small-world neuronal network research.
• Izhikevich and Rulkov models offer distinct advantages for simulating neuronal behavior and network construction.

Purpose of the Study:

• To investigate collective behaviors in small-world neuronal networks using Izhikevich and Rulkov models.
• To evaluate the computational efficiency of these models, particularly with GPU acceleration.

Main Methods:

• Simulated small-world neuronal networks based on mammalian cortex anatomy.
• Employed both the Izhikevich and Rulkov neuron models.
• Varied network parameters like connection probability and nearest neighbors.
• Utilized Graphics Processing Units (GPU) for accelerated computation.

Main Results:

• Coupled neurons exhibited diverse temporal and spatial characteristics based on parameter variations.
• GPU acceleration demonstrated significant speedup over CPU, increasing with neuron number and iterations.
• Both models proved suitable for constructing large-scale neural networks.

Conclusions:

• Small-world neuronal network models, like Izhikevich and Rulkov, can reproduce complex neural dynamics.
• GPU acceleration is crucial for simulating large-scale biological neural networks efficiently.
• This approach offers new possibilities for studying realistic neural systems.