Digital Multiplierless Realization of Coupled Wilson Neuron Model

Area of Science:

• Neuroscience
• Computer Engineering
• Computational Biology

Background:

• The human brain comprises 10^11 neurons with millisecond switching speeds, making neural networks a key area of neuromorphic research.
• Studying spiking neural networks aids in understanding brain function, diseases, and developing bio-mimetic processors and intelligent robots.
• Neuromorphic research, focusing on brain-inspired computing, is increasingly popular due to its potential for advanced AI and robotics.

Purpose of the Study:

• To implement an approximated Hodgkin-Huxley biological neuron model using the Wilson model.
• To digitally realize the modified Wilson neuron model for efficient hardware implementation.
• To evaluate the model's accuracy in reproducing biological neuron dynamics and its performance improvements.

Main Methods:

• Implemented the Wilson neuron model as an approximation of the Hodgkin-Huxley model.
• Modified the original Wilson model for efficient digital realization on hardware platforms.
• Performed hardware implementation on a field-programmable gate array (FPGA) to assess performance and accuracy.

Main Results:

• The proposed modified Wilson model successfully reproduced neuron dynamical behaviors.
• Hardware implementation on FPGA demonstrated that the modifications accurately imitate biological neuron behavior.
• The modified model achieved a 15% speed-up compared to the original Wilson model.
• Quantitative error metrics (mean normalized RMSE, RMSE, MAE) were 6.43, 0.44, and 0.31, respectively, indicating high fidelity.

Conclusions:

• The modified Wilson neuron model offers a feasible, low-cost, and efficient approach for digital realization of biological neuron dynamics.
• This optimized model is suitable for neuromorphic applications requiring accurate simulation of neural behavior.
• The FPGA implementation validates the model's effectiveness in mimicking biological neurons with enhanced performance.