We report an experimental implementation of width-tunable neurons to train a binary neural network. The angle-dependent magnetic behavior in an oxide thin film highly mimics neurons with width-controllable activation window, providing an opportunity to train the activation functions and weights toward binary values. We apply this feature to train the MNIST dataset using a 684-800-10 fully connected network and achieve a high accuracy of 97.4%, thus opening an implementation strategy toward training neural networks.
References
1.
J.
Schmidhuber
, Neural Networks
61
, 85
–117
(2015
).2.
D.
Hassabis
, D.
Kumaran
, C.
Summerfield
, and M.
Botvinick
, Neuron
95
(2
), 245
(2017
).3.
C. M.
Kim
, K. H.
Choi
, and Y. B.
Cho
, in Proceedings of the International Joint Conference on Neural Networks 2003
(IEEE
, 2003
), Vols. 1–4
, pp. 953
.4.
5.
H.
Qin
, R.
Gong
, X.
Liu
, X.
Bai
, J.
Song
, and N.
Sebe
, Pattern Recognition
105
, 107281
(2020
).6.
R.
Gong
, X.
Liu
, S.
Jiang
, T.
Li
, P.
Hu
, J.
Lin
, F.
Yu
, and J.
Yan
, in 2019 IEEE/CVF International Conference on Computer Vision (ICCV)
(IEEE
, 2019
), p. 4851
.7.
M.
Rastegari
, V.
Ordonez
, J.
Redmon
, A.
Farhadi
, in Computer Vision—ECCV 2016, Pt IV
, edited by B.
Leibe
, J.
Matas
, N.
Sebe
et al. (Springer
, 2016
), Vol. 9908
, p. 525
.8.
Z.
Wu
, D.
Lin
, X.
Tang
, D.
Soudry
, I.
Hubara
, R.
Meir
, Z.
Lin
, M.
Courbariaux
, R.
Memisevic
, and Y.
Bengio
, Adv. Neural Inf. Process. Syst.
2
, 963
(2015
).9.
T.
Simons
and D.-J.
Lee
, Electronics
8
(6
), 661
(2019
).10.
X. F.
Lin
, C.
Zhao
, and W.
Pan
, in Advances in Neural Information Processing Systems
, edited by I.
Guyon
, U. V.
Luxburg
, S.
Bengio
et al. (MIT Press
, 2017
), Vol. 30
.11.
12.
P. H.
Yin
, S.
Zhang
, J. C.
Lyu
, S.
Osher
, Y. Y.
Qi
, and J.
Xin
, Res. Math. Sci.
6
, 14
(2019
).13.
Y.
Zhao
and J.
Xiao
, Sci. Rep.
11
, 19797
(2021
).14.
See http://yann.lecun.com/exdb/mnist/ for “
MNIST
.”15.
F. L.
Lan
, H. Y.
Chen
, H. X.
Lin
, Y.
Bai
, Y.
Yu
, T.
Miao
, Y. Y.
Zhu
, T. Z.
Ward
, Z.
Gai
, W. B.
Wang
, L. F.
Yin
, E. W.
Plummer
, and J.
Shen
, Proc. Natl. Acad. Sci. U. S. A.
116
(10
), 4141
(2019
).16.
V.
Kalappattil
, R.
Geng
, S. H.
Liang
, D.
Mukherjee
, J.
Devkota
, A.
Roy
, M. H.
Luong
, N. D.
Lai
, L. A.
Hornak
, and T. D.
Nguyen
, J. Sci. Adv. Mater. Devices
2
(3
), 378
–384
(2017
).17.
J. S.
Lee
, D. A.
Arena
, P.
Yu
, C. S.
Nelson
, R.
Fan
, C. J.
Kinane
, S.
Langridge
, M. D.
Rossell
, R.
Ramesh
, and C. C.
Kao
, Phys. Rev. Lett.
105
(25
), 257204
(2010
).18.
S.
Oh
, Y. H.
Shi
, J.
del Valle
, P.
Salev
, Y. C.
Lu
, Z. S.
Huang
, Y.
Kalcheim
, I. K.
Schuller
, and D.
Kuzum
, Nat. Nanotechnol.
16
(6
), 680
(2021
).19.
N.
Srivastava
, Master's thesis, University of Toronto
, 2013
.20.
N.
Srivastava
, G.
Hinton
, A.
Krizhevsky
, I.
Sutskever
, and R.
Salakhutdinov
, J. Mach. Learn. Res.
15
, 1929
(2014
).21.
P. Y.
Simard
, D.
Steinkraus
, and J. C.
Platt
, in Seventh International Conference on Document Analysis and Recognition
(IEEE
, 2003
), Vols. I and II
, p. 958
.22.
See https://www.tensorflow.org/ for “
TensorFlow
.”23.
T.
Nomura
, K.
Yokoyama
, and S.
Nakagawa
, SMPTE Motion Imaging J. Kimoto
96
(11
), 1062
(1987
).24.
T. D.
Goodman
, and M. Appl. Opt. Mansuripur
35
(7
), 1107
(1996
).25.
See
S. D.
Bader
, J. Magn. Mater.
100
, 440
(1991
).26.
X.
Fan
, H.
Celik
, J.
Wu
et al., Nat Commun
5
, 3042
(2014
).27.
F.
Huang
, M. T.
Kief
, and G. J.
Mankey
, Phys. Rev. B
49
(6
), 3962
(1994
).28.
H.
Molegraaf
, J.
Hoffman
, C.
Vaz
et al., Adv. Mater.
21
(34
), 3470
–3474
(2009
).29.
Z. Q.
Qiu
and J.
Pearson
, Phys. Rev. Lett.
70
(7
), 1006
(1993
).30.
H.
Nishikawa
, E.
Houwman
, H.
Boschker
, M.
Mathews
, D. H. A.
Blank
, and G.
Rijnders
, Appl. Phys. Lett.
94
(4
), 042502
(2009
).31.
V. V.
Demidov
, G. A.
Ovsyannikov
, A. M.
Petrzhik
, I. V.
Borisenko
, A. V.
Shadrin
, and R.
Gunnarsson
, J. Appl. Phys.
113
(16
), 163909
(2013
).© 2021 Author(s). Published under an exclusive license by AIP Publishing.
2021
Author(s)
You do not currently have access to this content.
