INTERDISCIPLINARY PHYSICS AND RELATED AREAS OF SCIENCE AND TECHNOLOGY |
Prev
Next
|
|
|
A progressive surrogate gradient learning for memristive spiking neural network |
Shu Wang(王姝), Tao Chen(陈涛), Yu Gong(龚钰), Fan Sun(孙帆), Si-Yuan Shen(申思远), Shu-Kai Duan(段书凯)†, and Li-Dan Wang(王丽丹) |
College of Artificial Intelligence, Southwest University, Chongqing 400715, China |
|
|
Abstract In recent years, spiking neural networks (SNNs) have received increasing attention of research in the field of artificial intelligence due to their high biological plausibility, low energy consumption, and abundant spatio-temporal information. However, the non-differential spike activity makes SNNs more difficult to train in supervised training. Most existing methods focusing on introducing an approximated derivative to replace it, while they are often based on static surrogate functions. In this paper, we propose a progressive surrogate gradient learning for backpropagation of SNNs, which is able to approximate the step function gradually and to reduce information loss. Furthermore, memristor cross arrays are used for speeding up calculation and reducing system energy consumption for their hardware advantage. The proposed algorithm is evaluated on both static and neuromorphic datasets using fully connected and convolutional network architecture, and the experimental results indicate that our approach has a high performance compared with previous research.
|
Received: 13 November 2022
Revised: 07 January 2023
Accepted manuscript online: 08 February 2023
|
PACS:
|
87.19.L-
|
(Neuroscience)
|
|
87.19.ll
|
(Models of single neurons and networks)
|
|
89.20.Ff
|
(Computer science and technology)
|
|
Fund: Project supported by the Natural Science Foundation of Chongqing (Grant No. cstc2021jcyj-msxmX0565), the Fundamental Research Funds for the Central Universities (Grant No. SWU021002), and the Graduate Research Innovation Project of Chongqing (Grant No. CYS22242). |
Corresponding Authors:
Shu-Kai Duan
E-mail: duansk@swu.edu.cn
|
Cite this article:
Shu Wang(王姝), Tao Chen(陈涛), Yu Gong(龚钰), Fan Sun(孙帆), Si-Yuan Shen(申思远), Shu-Kai Duan(段书凯), and Li-Dan Wang(王丽丹) A progressive surrogate gradient learning for memristive spiking neural network 2023 Chin. Phys. B 32 068704
|
[1] Vaila R, Chiasson J and Saxena V2020 arXiv:2002.11843 [cs.NE] [2] Gast R, Schmidt H and Knösche T R2020 Neural Comput. 32 1615 [3] Maass W, Natschläger T and Markram H2002 Neural Comput. 14 2531 [4] Maass W1997 Neural Networks 10 1659 [5] Hodgkin A L and Huxley A F1952 J. Physiol. 117 500 [6] Vreeken J2003 Spiking neural networks, an introduction [7] Cortex I T 2005 Science 310 863 [8] Merolla P A, Arthur J V and Alvarez-Icaza R2014 Science 345 668 [9] Davies M, Srinivasa N and Lin T H2018 IEEE Micro 38 82 [10] Prezioso M, Merrikh-Bayat F and Hoskins B D2015 Nature 52 61 [11] Hunsberger E and Eliasmith C2015 arXiv:1510.08829 [cs.LG] [12] Neil D and Liu S C 2016 IEEE International Symposium on Circuits and Systems (ISCAS), May 22-25, 2016, Montreal, QC, Canada, pp. 2282-2285 [13] Rueckauer B, Lungu I A and Hu Y2017 Front. Neurosci. 11 682 [14] Deng S and Gu S2021 arXiv:2103.00476 [cs.NE] [15] Fang W, Yu Z and Chen Y2021 arXiv:2007.05785 [cs.NE] [16] Kelso S R, Ganong A H and Brown T H1986 Proc. Natl. Acad. Sci. USA 83 5326 [17] Hebb D O 2005 The organization of behavior: A neuropsychological theory (New York: Psychology Press) [18] Querlioz D, Bichler Oa and Dollfus P2013 IEEE Trans. Nanotechnol. 12 288 [19] Diehl P U and Cook M2015 Front. Comput. Neurosci. 9 99 [20] Guo S, Yu Z and Deng F 2017 IEEE Trans. Cybernet. 49 133 [21] Kheradpisheh S R, Ganjtabesh M and Masquelier T2016 Neurocomputing 205 382 [22] Mohemmed A, Schliebs S and Matsuda S2012 Int. J. Neural Syst. 22 1250012 [23] Bohte S M, Kok J N and La Poutré J A 2000 SpikeProp: Backpropagation for networks of spiking neurons (Belgium: The European Symposium on Artificial Neural Networks) 48 419 [24] Gütig R and Sompolinsky H2006 Nat. Neurosci. 9 420 [25] Ponulak F and Kasiński A2010 Neural Comput. 22 467 [26] Lee J H, Delbruck T and Pfeiffer M2016 Front. Neurosci. 10 508 [27] Jin Y, Zhang W and Li P2018 arXiv:1805.07866 [cs.NE] [28] Wu Y, Deng L and Li G 2018 Front. Neurosci. 12 331 [29] Neftci E O, Mostafa H and Zenke F 2019 IEEE Signal Proc. Mag. 36 61 [30] Chua L.1971 IEEE Trans. Circuit Theory 18 507 [31] Li Y, Zhong Y and Zhang J2014 Sci. Rep. 4 4906 [32] Jo S H, Chang T and Ebong I2010 Nano Lett. 10 1297 [33] Yao P, Wu H and Gao B2017 Nat. Commun. 8 1 [34] Li J, Zhou G and Li Y2022 Artif Intell Rev 55 657 [35] Yao P, Wu H and Gao B2020 Nature 577 641 [36] Kheradpisheh S R and Masquelier T2020 Int. J. Neural Syst. 30 2050027 [37] Li Y, Guo Y and Zhang S 2021 Adv. Neural Inf. Process Syst. 34 23426 [38] Vaezi M S and Vaezi A2018 arXiv:1810.00864 [cond-mat.str-el] [39] Zenke F and Ganguli S2018 Neural Comput. 30 1514 [40] Bellec G, Salaj D, Subramoney A et al.2018 arXiv:1803.09574 [cs.NE] [41] Esser S K, Merolla P A and Arthur J V2016 Proc. Natl. Acad. Sci. USA 113 11441 [42] Bohte S M 2011 International Conference on Artificial Neural Networks pp. 60-68. Lecture Notes in Computer Science Vol. 6791 (Berlin: Springer) [43] Hodgkin A L and Huxley A F1952 J. Physiol. 117 500 [44] Izhikevich E M2003 IEEE Trans. Neural Netw. 14 1569 [45] Liu Y H and Wang X J2001 J. Comput. Neurosci. 10 25 [46] Strukov D B, Snider G S and Stewart D R2008 Nature 453 80 [47] Abdalla H and Pickett M D 2011 IEEE International Symposium of Circuits and Systems (ISCAS), May 15-18, 2011, Rio de Janeiro, Brazil, pp. 1832-1835 [48] Kvatinsky S, Friedman E G and Kolodny A2012 IEEE Trans. Circuits Syst. I: Regular Papers 60 211 [49] Chen L, Gong C and Li C2021 Sustain Cities Soc. 69 102849 [50] Zhou W, Wen S and Liu Y2023 Neural Netw. 158 293 [51] Lee C, Sarwar S S and Panda P2020 Front. Neurosci. 14 119 [52] Rueckauer B, Lungu I A and Hu Y2017 Front. Neurosci. 11 682 [53] Zhang W and Li P2019 arXiv:1908.06378 [cs.NE] [54] Zhang W and Li P2020 arXiv:2002.10085 [cs.NE] [55] Xiao H, Rasul K and Vollgraf R2017 arXiv:1708.07747 [cs.LG] [56] He W, Wu Y J and Deng L2020 Neural Networks 132 108 [57] Xing Y, Di Caterina G and Soraghan J2020 Front. Neurosci. 14 1143 [58] He W, Wu Y J and Deng L2020 Neural Netw. 132 108 [59] Kaiser J, Mostafa H and Neftci E2020 Front. Neurosci. 14 424 |
No Suggested Reading articles found! |
|
|
Viewed |
|
|
|
Full text
|
|
|
|
|
Abstract
|
|
|
|
|
Cited |
|
|
|
|
Altmetric
|
blogs
Facebook pages
Wikipedia page
Google+ users
|
Online attention
Altmetric calculates a score based on the online attention an article receives. Each coloured thread in the circle represents a different type of online attention. The number in the centre is the Altmetric score. Social media and mainstream news media are the main sources that calculate the score. Reference managers such as Mendeley are also tracked but do not contribute to the score. Older articles often score higher because they have had more time to get noticed. To account for this, Altmetric has included the context data for other articles of a similar age.
View more on Altmetrics
|
|
|