Investigation on the Wilson Neuronal Model: Optimized Approximation and Digital Multiplierless Implementation.

Area of Science:

• Neuromorphic Engineering
• Computational Neuroscience
• Digital Hardware Design

Background:

• Neurons are fundamental units in the Central Nervous System (CNS), responsible for receiving, processing, and transmitting data.
• Neuromorphic engineering integrates physics, mathematics, and electronics to mimic brain functions.
• Existing neuron models can be computationally intensive and resource-heavy for digital implementation.

Purpose of the Study:

• To present a modified Wilson Neuron Model (WMN) optimized for multiplierless digital realization.
• To reduce overhead costs and increase system frequency in neuromorphic hardware.
• To validate the WMN model's performance and resource efficiency on FPGA hardware.

Main Methods:

• The Wilson Multiplierless Neuron (WMN) model utilizes power-2 functions, Look-Up Table (LUT) approach, and shifters.
• Multiplierless digital realization techniques are employed for efficient hardware implementation.
• The proposed model's spiking patterns and dynamical pathways are designed to emulate the original Wilson neuron model.

Main Results:

• The WMN model achieved a system speed-up of 210 MHz, significantly higher than the original model's 85 MHz.
• Overall FPGA resource saving for the WMN model is 96.86%, compared to 95.13% for the original model.
• Hardware validation on a Xilinx Virtex II FPGA demonstrated increased system frequency and reduced resource utilization.

Conclusions:

• The WMN model offers a significant improvement in speed and resource efficiency for digital neuromorphic systems.
• The model maintains the spiking patterns and dynamical pathways of the original Wilson neuron.
• The WMN model provides a viable platform for simulating neural networks and investigating neurological diseases like Epilepsy.