Real-time million-synapse simulation of rat barrel cortex.

Area of Science:

• Computational Neuroscience
• Computer Architecture
• Neuroscience

Background:

• Neural circuit simulations are limited by computational resources and differences between brain architecture and high-performance computers.
• Existing computing systems struggle to emulate the brain's parallelism and communication patterns effectively.

Purpose of the Study:

• To demonstrate the capability of the SpiNNaker computer architecture for large-scale neural circuit simulations.
• To validate SpiNNaker's performance in emulating biological neural tissue structure and function.

Main Methods:

• Utilized thousand-processor SpiNNaker prototypes to simulate rodent barrel cortex models.
• Employed the PyNN library for model specification and Python for experimental control and analysis.
• Modeled neural circuits comprising 50,000 neurons and 50 million synapses.

Main Results:

• Successfully simulated a detailed model of the rodent barrel system.
• Reproduced known thalamocortical response transformations and balanced excitation-inhibition dynamics.
• Observed accurate spatiotemporal spread of activity across simulated cortical layers.

Conclusions:

• SpiNNaker prototypes show significant progress towards simulating entire cortical areas.
• The architecture effectively emulates neural tissue, overcoming limitations of traditional computing.
• This work paves the way for future large-scale neural simulations on million-processor SpiNNaker systems.