Digital System Implementation and Large-Scale Approach in Neuronal Modeling Using Adex Biological Neuron
Muhammad Akmal Chaudhary, Fawwaz Hazzazi, Milad Ghanbarpour
Abstract
The Adex neuron model is implemented in a system using the high-matching method under the name Digital-High-Matching Adex Neuron (D-HMAN). Comparing this model to all other comparable efforts, it has reduced implementation costs while accurately reproducing various spiking characteristics, similar to those of biological neurons. High computational performance, minimal hardware cost, and strong resemblance to the original model are all evidenced by experimental findings. In comparison to the original Adex neuron, the suggested D-HMAN version on FPGA uses a lot less hardware and also higher levels of speed and frequency (618.24 vs. 348.71 MHz in Zynq board). With this update, the original model’s temporal patterns of neuron firing and the ever-changing nature of neural activity are replicated by performing computations using affordable fixed-point calculations. For large-scale neuromorphic design aiming for low-cost hardware realization, this makes the suggested model an attractive contender. The suggested model is implemented on the large-scale level in a digital form as a case study.