journal article·2025-03Open access

Ultrafast Probabilistic Neuron in an Artificial Spin Ice for Robust Deep Neural Networks

Zhongyu Liang, Tong Bu, Zijian Lyu, Zhentao Liu, Aleš Hrabec, Leran Wang, Yankun Dou, Jianhao Ding, Peipei Ge, Wenyun Yang, Tiejun Huang, Jinbo Yang, Laura J. Heyderman, Yunquan Liu, Zhaofei Yu, Zhaochu Luo

Advanced Functional Materials· Volume 35· Issue 11

Abstract

Abstract Deep neural networks (DNNs) have proved to be remarkably successful in various domains, in particular for implementing complex functions and performing sophisticated tasks. However, their vulnerability to adversarial noise undermines their reliability for safety‐critical tasks. Despite attempts to improve the robustness using algorithmic approaches, an effective hardware implementation is still lacking. Here an artificial probabilistic neuron device is proposed based on arrays of coupled nanomagnets, referred to as artificial spin ices, which return a nonlinear function with built‐in stochasticity in response to an ultrafast laser‐induced excitation. By exploiting solid‐state ionic gating, the magnetic coupling is electrically modulated, as a result of the magnetic anisotropy‐mediated competition of the symmetric exchange interaction and Dzyaloshinskii‐Moriya interaction, and hence the stochastic property of the neuron device at runtime can be tuned. Stochastic DNNs are then constructed with an output layer comprising several probabilistic neuron devices. Compared to conventional DNNs, the stochastic DNNs exhibit an order of magnitude greater resistance to adversarial noise, providing a significant improvement in robustness. This approach opens the way to more secure and reliable DNNs, enabling broader uses in real‐world applications.
Source: crossref · Quality: baseline

Research context

OpenAlex topics: Neural Networks and Reservoir Computing · Advanced Memory and Neural Computing · Ferroelectric and Negative Capacitance Devices

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

State Key Laboratory of Artificial Microstructure and Mesoscopic Physics School of Physics Peking University Beijing 100871 China

School of Advanced Manufacturing Fuzhou University Jinjiang 362200 China

Institute for Artificial Intelligence Peking University Beijing 100871 China

School of Computer Science Peking University Beijing 100871 China

Laboratory for Mesoscopic Systems Department of Materials ETH Zurich Zurich 8093 Switzerland

PSI Center for Neutron and Muon Sciences Villigen PSI 5232 Switzerland

Collaborative Innovation Center of Extreme Optics Shanxi University Taiyuan 030006 China

Funding

National Key Research and Development Program of China · National Natural Science Foundation of China · Fundamental Research Funds for the Central Universities · Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung · Beijing Nova Program

References (49)

Show source-provided references
1. 10.1038/nature1453910.1038/nature14539 ↗
2. 10.1038/s41586-021-03819-210.1038/s41586-021-03819-2 ↗
3. 10.1038/s41586-023-06004-910.1038/s41586-023-06004-9 ↗
4. C.Szegedy W.Zaremba I.Sutskever J.Bruna D.Erhan I.Goodfellow R.Fergus presented atInt. Conf. Learn. Rep. Banff Canada April2014.
5. I. J.Goodfellow J.Shlens C.Szegedy presented atInt. Conf. Learn. Rep. San Diego USA May2015.
6. K.Eykholt I.Evtimov E.Fernandes B.Li A.Rahmati C.Xiao A.Prakash T.Kohno D.Song presented atIEEE Conf. Comput. Vision Pattern Recognit. Salt Lake City USA June2018.
7. X.Liu M.Cheng H.Zhang C.Hsieh presented atEur. Conf. Comput. Vision Munich Germany September2018.
8. Z.He A.Siraj Rakin D.Fan presented atIEEE Conf. Comput. Vision Pattern Recognit. California USA June2019.
9. 10.1126/science.aaw439910.1126/science.aaw4399 ↗
10. 10.1002/adma.20220456910.1002/adma.202204569 ↗
11. 10.1038/nrn225810.1038/nrn2258 ↗
12. 10.1038/s41467-019-12035-610.1038/s41467-019-12035-6 ↗
13. 10.1038/s41467-019-13103-710.1038/s41467-019-13103-7 ↗
14. 10.1038/s41467-022-30305-810.1038/s41467-022-30305-8 ↗
15. 10.1038/nnano.2016.7010.1038/nnano.2016.70 ↗
16. 10.1103/RevModPhys.85.147310.1103/RevModPhys.85.1473 ↗
17. 10.1002/smll.20200314110.1002/smll.202003141 ↗
18. 10.1038/s42254-019-0118-310.1038/s42254-019-0118-3 ↗
19. 10.1002/advs.20240218210.1002/advs.202402182 ↗
20. 10.1038/s41586-019-1557-910.1038/s41586-019-1557-9 ↗
21. 10.1103/PhysRevApplied.17.02405210.1103/PhysRevApplied.17.024052 ↗
22. J. H.Jensen E.Folven G.Tufte presented atConf. Artif. Life Tokyo Japan July2018.
23. 10.1038/s41565-022-01091-710.1038/s41565-022-01091-7 ↗
24. 10.1038/s41467-023-38286-y10.1038/s41467-023-38286-y ↗
25. 10.1088/1741-2560/13/5/05100110.1088/1741-2560/13/5/051001 ↗
26. 10.1038/nphys261310.1038/nphys2613 ↗
27. 10.1126/science.aau791310.1126/science.aau7913 ↗
28. 10.1038/s41467-023-41830-510.1038/s41467-023-41830-5 ↗
29. 10.1038/s41586-020-2061-y10.1038/s41586-020-2061-y ↗
30. 10.1038/s41563-018-0211-510.1038/s41563-018-0211-5 ↗
31. 10.1103/PhysRevLett.76.425010.1103/PhysRevLett.76.4250 ↗
32. 10.1103/RevModPhys.82.273110.1103/RevModPhys.82.2731 ↗
33. 10.1038/ncomms210810.1038/ncomms2108 ↗
34. 10.1002/admi.20220128310.1002/admi.202201283 ↗
35. 10.1007/BF0175905410.1007/BF01759054 ↗
36. 10.1049/ip-vis:2003036210.1049/ip-vis:20030362 ↗
37. 10.1016/j.neuroimage.2012.04.06110.1016/j.neuroimage.2012.04.061 ↗
38. 10.1109/TNNLS.2018.287456510.1109/TNNLS.2018.2874565 ↗
39. 10.1038/ncomms1027510.1038/ncomms10275 ↗
40. 10.1038/s41928-021-00593-x10.1038/s41928-021-00593-x ↗
41. 10.1038/s41566-020-00754-y10.1038/s41566-020-00754-y ↗
42. 10.1002/adma.20221021610.1002/adma.202210216 ↗
43. 10.1038/s41586-022-04714-010.1038/s41586-022-04714-0 ↗
44. 10.1126/science.abh207610.1126/science.abh2076 ↗
45. 10.1103/PhysRevLett.99.04760110.1103/PhysRevLett.99.047601 ↗
46. 10.1126/science.125349310.1126/science.1253493 ↗
47. Krizhevsky A. · Adv. Neural Inf. Process. Syst. · 2012 · 25 · 1090
48. 10.1038/323533a010.1038/323533a0 ↗
49. I.Loshchilov F.Hutter presented atInt. Conf. Learn. Rep. Toulon France April2017.