FPGA implementation of a biological neural network based on the Hodgkin-Huxley neuron model

Abstract

A set of techniques for efficient implementation of Hodgkin-Huxley-based (H-H) model of a neural network on FPGA (Field Programmable Gate Array) is presented. The central implementation challenge is H-H model complexity that puts limits on the network size and on the execution speed. However, basics of the original model cannot be compromised when effect of synaptic specifications on the network behavior is the subject of study. To solve the problem, we used computational techniques such as CORDIC (Coordinate Rotation Digital Computer) algorithm and step-by-step integration in the implementation of arithmetic circuits. In addition, we employed different techniques such as sharing resources to preserve the details of model as well as increasing the network size in addition to keeping the network execution speed close to real time while having high precision. Implementation of a two mini-columns network with 120/30 excitatory/inhibitory neurons is provided to investigate the characteristic of our method in practice. The implementation techniques provide an opportunity to construct large FPGA-based network models to investigate the effect of different neurophysiological mechanisms, like voltage-gated channels and synaptic activities, on the behavior of a neural network in an appropriate execution time. Additional to inherent properties of FPGA, like parallelism and re-configurability, our approach makes the FPGA-based system a proper candidate for study on neural control of cognitive robots and systems as well.

A Scalable FPGA Architecture for Randomly Connected Networks of Hodgkin-Huxley Neurons

“…Our results confirm that our approximation techniques dramatically reduce both computational complexities and memory usage and make our architecture suitable for embedded applications, even though the most complex neuron model is used in our work. While some works exploit either Hodgkin-Huxley neuron model or its variants (Zhang et al, 2013 ; Smaragdos et al, 2014 ; Osorio, 2016 ; Yang et al, 2017 ), some other works utilize its reduced forms (Graas et al, 2004 ; Yaghini Bonabi et al, 2014 ). Meanwhile, the majority of researchers use simple neuron models such as Izhikevich or leaky integrate and fire (Cassidy et al, 2007 ; Soleimani et al, 2012 ; Ambroise et al, 2013 ; Furber et al, 2013 ; Cheung et al, 2016 ; Pani et al, 2017 ) to cope with the computational complexity of the HH model.…”

“…Although PWL methods for approximating nonlinear functions are widely used in the literature, some works focus on other realizations of nonlinear functions. For instance, in Yaghini Bonabi et al ( 2014 ), the CORDIC is used to implement exponential functions of the HH neuron model. The CORDIC is more accurate than a PWL approximation at the cost of more resource usage.…”

“…Although all of them show reasonable accuracy under certain circumstances, their complexities make it difficult to have a small and fast hardware implementation. Iterative techniques such as CORDIC approach reduce the hardware size, but its iterative nature causes more latency than the noniterative algorithms (Parhami, 2010 ; Yaghini Bonabi et al, 2014 ).…”

“…Some research projects exploit FPGAs in the heart of their SNN. For example, in Yaghini Bonabi et al ( 2014 ), the network consists of two competitive mini-columns of HH neuron model. The final design is implemented on a Xilinx Virtex 7 FPGA.…”

Digital FPGA implementation of spontaneous astrocyte signalling

“…This approach is inspired by information processing in biological neural networks and makes them interesting choice for the efficient realization of biological neural networks . From the neuromorphic engineering point of view, hardware implementation of different parts of the central nervous system (CNS), including neurons, synapses, and astrocytes, is significant in this field …”