An FPGA Implementation of Convolutional Spiking Neural Networks for Radioisotope Identification

Mike Huang; Edward Jones; Siru Zhang; Shouyu Xie; Steve Furber; Yannis Goulermas; Edward Marsden; Ian Baistow; Srinjoy Mitra; Alister Hamilton

doi:10.1109/ISCAS51556.2021.9401412

An FPGA Implementation of Convolutional Spiking Neural Networks for Radioisotope Identification

Mike Huang, Edward Jones, Siru Zhang, Shouyu Xie, Steve Furber, Yannis Goulermas, Edward Marsden, Ian Baistow, Srinjoy Mitra, Alister Hamilton

School of Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

This paper presents a detailed FPGA implementation methodology of Convolutional Spiking Neural Network based low-power and high-resolution radioisotope identification. A power budget of 74 mW has been achieved on an FPGA with the inference accuracy of 90.62% at a synthetic dataset. The design verification and chip validation methods are presented. It also discusses SNN simulation on SpiNNaker for rapid prototyping and various considerations specific to the application such as test distance, integration time, and SNN hyperparameter selections.

Original language	English
Title of host publication	2021 IEEE International Symposium on Circuits and Systems (ISCAS)
Publisher	Institute of Electrical and Electronics Engineers
ISBN (Print)	978-1-7281-9201-7
DOIs	https://doi.org/10.1109/ISCAS51556.2021.9401412
Publication status	Published - 27 Apr 2021

Access to Document

10.1109/ISCAS51556.2021.9401412

https://arxiv.org/abs/2102.12565

Cite this

Huang, M., Jones, E., Zhang, S., Xie, S., Furber, S., Goulermas, Y., Marsden, E., Baistow, I., Mitra, S., & Hamilton, A. (2021). An FPGA Implementation of Convolutional Spiking Neural Networks for Radioisotope Identification. In 2021 IEEE International Symposium on Circuits and Systems (ISCAS) Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/ISCAS51556.2021.9401412

@inproceedings{a1b3487240e045db835910a35d18799a,

title = "An FPGA Implementation of Convolutional Spiking Neural Networks for Radioisotope Identification",

abstract = "This paper presents a detailed FPGA implementation methodology of Convolutional Spiking Neural Network based low-power and high-resolution radioisotope identification. A power budget of 74 mW has been achieved on an FPGA with the inference accuracy of 90.62% at a synthetic dataset. The design verification and chip validation methods are presented. It also discusses SNN simulation on SpiNNaker for rapid prototyping and various considerations specific to the application such as test distance, integration time, and SNN hyperparameter selections. ",

author = "Mike Huang and Edward Jones and Siru Zhang and Shouyu Xie and Steve Furber and Yannis Goulermas and Edward Marsden and Ian Baistow and Srinjoy Mitra and Alister Hamilton",

year = "2021",

month = apr,

day = "27",

doi = "10.1109/ISCAS51556.2021.9401412",

language = "English",

isbn = "978-1-7281-9201-7",

booktitle = "2021 IEEE International Symposium on Circuits and Systems (ISCAS)",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

}

Huang, M, Jones, E, Zhang, S, Xie, S, Furber, S, Goulermas, Y, Marsden, E, Baistow, I, Mitra, S & Hamilton, A 2021, An FPGA Implementation of Convolutional Spiking Neural Networks for Radioisotope Identification. in 2021 IEEE International Symposium on Circuits and Systems (ISCAS). Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/ISCAS51556.2021.9401412

TY - GEN

T1 - An FPGA Implementation of Convolutional Spiking Neural Networks for Radioisotope Identification

AU - Huang, Mike

AU - Jones, Edward

AU - Zhang, Siru

AU - Xie, Shouyu

AU - Furber, Steve

AU - Goulermas, Yannis

AU - Marsden, Edward

AU - Baistow, Ian

AU - Mitra, Srinjoy

AU - Hamilton, Alister

PY - 2021/4/27

Y1 - 2021/4/27

N2 - This paper presents a detailed FPGA implementation methodology of Convolutional Spiking Neural Network based low-power and high-resolution radioisotope identification. A power budget of 74 mW has been achieved on an FPGA with the inference accuracy of 90.62% at a synthetic dataset. The design verification and chip validation methods are presented. It also discusses SNN simulation on SpiNNaker for rapid prototyping and various considerations specific to the application such as test distance, integration time, and SNN hyperparameter selections.

AB - This paper presents a detailed FPGA implementation methodology of Convolutional Spiking Neural Network based low-power and high-resolution radioisotope identification. A power budget of 74 mW has been achieved on an FPGA with the inference accuracy of 90.62% at a synthetic dataset. The design verification and chip validation methods are presented. It also discusses SNN simulation on SpiNNaker for rapid prototyping and various considerations specific to the application such as test distance, integration time, and SNN hyperparameter selections.

U2 - 10.1109/ISCAS51556.2021.9401412

DO - 10.1109/ISCAS51556.2021.9401412

M3 - Conference contribution

SN - 978-1-7281-9201-7

BT - 2021 IEEE International Symposium on Circuits and Systems (ISCAS)

PB - Institute of Electrical and Electronics Engineers

ER -

An FPGA Implementation of Convolutional Spiking Neural Networks for Radioisotope Identification

Abstract

Access to Document

Fingerprint

Cite this