Toward the Optimal Design and FPGA Implementation of Spiking Neural Networks

The performance of a biologically plausible spiking neural network (SNN) largely depends on the model parameters and neural dynamics. This article proposes a parameter optimization scheme for improving the performance of a biologically plausible SNN and a parallel on-field-programmable gate array (F...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transaction on neural networks and learning systems Vol. 33; no. 8; pp. 3988 - 4002
Main Authors: Guo, Wenzhe, Yantir, Hasan Erdem, Fouda, Mohammed E., Eltawil, Ahmed M., Salama, Khaled Nabil
Format: Journal Article
Language:English
Published: United States IEEE 01.08.2022
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:
Biological system modeling
Biology
Central processing units
Computational modeling
CPUs
Energy consumption
Euler method
Field programmable gate arrays
field-programmable gate array (FPGA) platform
Firing pattern
Inference
Information processing
Machine learning
Mathematical models
Neural networks
neuromorphic computing
Neurons
Numerical methods
Numerical models
on-chip learning
Online instruction
Optimization
parallel architecture
parameter optimization
Parameters
Runge-Kutta (RK) method
Runge-Kutta method
Simulation
Spiking
spiking neural network (SNN)
Training
unsupervised spike-timing-dependent plasticity (STDP) learning
ISSN:2162-237X, 2162-2388, 2162-2388
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The performance of a biologically plausible spiking neural network (SNN) largely depends on the model parameters and neural dynamics. This article proposes a parameter optimization scheme for improving the performance of a biologically plausible SNN and a parallel on-field-programmable gate array (FPGA) online learning neuromorphic platform for the digital implementation based on two numerical methods, namely, the Euler and third-order Runge-Kutta (RK3) methods. The optimization scheme explores the impact of biological time constants on information transmission in the SNN and improves the convergence rate of the SNN on digit recognition with a suitable choice of the time constants. The parallel digital implementation leads to a significant speedup over software simulation on a general-purpose CPU. The parallel implementation with the Euler method enables around <inline-formula> <tex-math notation="LaTeX">180\times </tex-math></inline-formula> (<inline-formula> <tex-math notation="LaTeX">20\times </tex-math></inline-formula>) training (inference) speedup over a Pytorch-based SNN simulation on CPU. Moreover, compared with previous work, our parallel implementation shows more than <inline-formula> <tex-math notation="LaTeX">300\times </tex-math></inline-formula> (<inline-formula> <tex-math notation="LaTeX">240\times </tex-math></inline-formula>) improvement on speed and <inline-formula> <tex-math notation="LaTeX">180\times </tex-math></inline-formula> (<inline-formula> <tex-math notation="LaTeX">250\times </tex-math></inline-formula>) reduction in energy consumption for training (inference). In addition, due to the high-order accuracy, the RK3 method is demonstrated to gain <inline-formula> <tex-math notation="LaTeX">2\times </tex-math></inline-formula> training speedup over the Euler method, which makes it suitable for online training in real-time applications.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:2162-237X
2162-2388
2162-2388
DOI:10.1109/TNNLS.2021.3055421