A device-circuit-algorithm review of physics-driven platforms for next-generation artificial intelligence

Abstract

Recent advancements in artificial intelligence(AI) have significantly enhanced the perception and cognition of language and the decision-making process. However, the increase in computational power and energy efficiency has also led to the emergence of computational and energy needs. Traditional Von Neumann architectures, with the separation of memory and processing, are costly due to numerous data transfers between hierarchical memory systems. Therefore, in large-scale AI workloads, energy consumption and data transfer latency are major challenges to scalability and efficiency. Using physical phenomena to compute and combining devices, circuits, and algorithms by co-design, these issues are resolved by physics-driven computing. In-memory operations, nonlinearity, and parallelism are also enabled by resistive switching, spin dynamics, phase transitions, and optical interference at the device level, thereby significantly reducing data transmission and energy consumption. Unless the entire hardware stack, including devices, circuits, architectures, and learning algorithms, is coordinated, one achieves limited system-level benefits. This review presents a unified physics-based AI hardware design by jointly optimizing device, circuit, and algorithm. It also discusses new device platforms, including memristive, spintronic, phase-change, photonic, and neuromorphic complementary metal-oxide-semiconductor (CMOS), and how the physics underlying these principles can be used to implement computational primitives. It is discussed further at the circuit level, e.g., in-memory computing arrays, event cameras, neuromorphic processors, and algorithms such as hardware-conscious training, spike-based neural computation, phase-based training, and stochastic inference are also discussed in these physical resources. Together, these innovations enable the development of scalable, adaptable, and energy-efficient intelligent hardware through integrated co-design of materials, electronics, and machine learning.

Keywords

Physics-driven computing / Neuromorphic circuits / Device-circuit-algorithm co-design / Memristors / Spintronics / Phase change devices / Photonic artificial intelligence

References

[1] LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015; 521(7533):436-444. doi: 10.1038/nature14539

[2] Krizhevsky A, Sutskever I, Hinton GE. ImageNet classification with deep convolutional neural networks. Adv Neural Inf Process Syst. 2012; 25:1097-1105.

[3] Mnih V, Kavukcuoglu K, Silver D, et al. Human-level control through deep reinforcement learning. Nature. 2015; 518(7540):529-533. doi: 10.1038/nature14236

[4] Jouppi NP, Young C, Patil N, et al. In-datacenter performance analysis of a tensor processing unit. In:Proceedings of the 44th Annual International Symposium on Computer Architecture. 2017:1-12. doi: 10.1145/3079856.3080246

[5] Horowitz M. Computing’s energy problem (and what we can do about it). IEEE Int Solid-State Circuits Conf Dig Tech Pap. 2014:10-14. doi: 10.1109/ISSCC.2014.6757323

[6] Shalf J, Leland R. Computer. 2015; 48(12):14-23. doi: 10.1109/MC.2015.374

[7] Sebastian A, Le Gallo M, Khaddam-Aljameh R, Eleftheriou E. Memory devices and applications for in-memory computing. Nat Nanotechnol. 2020; 15(7):529-544. doi: 10.1038/s41565-020-0655-z

[8] Indiveri G, Liu SC. Memory and information processing in neuromorphic systems. Proc IEEE. 2015; 103(8):1379-1397. doi: 10.1109/JPROC.2015.2444094

[9] Burr GW, Shelby RM, Sebastian A, et al. Neuromorphic computing and engineering. Adv Phys X. 2017; 2(1):89-124. doi: 10.1080/23746149.2016.1259585

[10] Moitra A, Bhattacharjee A, Li Y, Kim Y, Panda P. When in-memory computing meets spiking neural networks-A perspective on device-circuit-system-and-algorithm co-design. Appl Phys Rev. 2024; 11(3):031325. doi: 10.1063/5.0211040

[11] Li C, Hu M, Li Y, et al. Analogue signal and image processing with large memristor crossbars. Nat Electron. 2018; 1:52-59. doi: 10.1038/s41928-017-0002-z

[12] Prezioso M, Merrikh-Bayat F, Hoskins BD, Adam GC, Likharev KK, Strukov DB. Training and operation of an integrated neuromorphic network based on metal-oxide memristors. Nature. 2015; 521(7550):61-64. doi: 10.1038/nature14441

[13] Ambrogio S, Narayanan P, Tsai H, et al. Equivalent-accuracy accelerated neural-network training using analogue memory. Nature. 2018; 558(7708):60-67. doi: 10.1038/s41586-018-0180-5

[14] Shen Y, Harris NC, Skirlo S, et al. Deep learning with coherent nanophotonic circuits. Nat Photonics. 2017; 11(7):441-446. doi: 10.1038/nphoton.2017.93

[15] Wang Z, Joshi S, Savel’ev SE, et al. Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing. Nat Mater. 2017; 16:101-108. doi: 10.1038/nmat4756

[16] Rasch MJ, Mackin C, Le Gallo M, et al. Hardware-aware training for large-scale and diverse deep learning inference workloads using in-memory computing-based accelerators. Nat Commun. 2023; 14:5282. doi: 10.1038/s41467-023-40770-4

[17] Wang Z, Wu H, Burr GW, et al. Resistive switching materials for information processing. Nat Rev Mater. 2020; 5(3):173-195. doi: 10.1038/s41578-019-0159-3

[18] Yu S. Neuro-inspired computing with emerging nonvolatile memories. Proc IEEE. 2018; 106(2):260-285. doi: 10.1109/JPROC.2018.2790840

[19] Feldmann J, Youngblood N, Karpov M, et al. Parallel convolutional processing using an integrated photonic tensor core. Nature. 2021; 589(7840):52-58. doi: 10.1038/s41586-020-03070-1

[20] Wulf WA, McKee SA. Hitting the memory wall: Implications of the obvious. ACM SIGARCH Comput Archit News. 1995; 23(1):20-24. doi: 10.1145/216585.216588

[21] Sze V, Chen YH, Yang TJ, Emer JS. Efficient processing of deep neural networks: A tutorial and survey. Proc IEEE. 2017; 105(12):2295-2329. doi: 10.1109/JPROC.2017.2761740

[22] Jouppi NP, Yoon DH, Ashcraft M, et al. Ten lessons from three generations shaped Google’s TPUv4i. In: Proceedings of the 2021 ACM/IEEE 48th Annu Int Symp Comput Archit (ISCA). 2021:1-14. doi: 10.1109/ISCA52012.2021.00010

[23] Chen YH, Krishna T, Emer JS, Sze V. Eyeriss: An energy-efficient reconfigurable accelerator for deep convolutional neural networks. IEEE J Solid-State Circuits. 2017; 52(1):127-138. doi: 10.1109/JSSC.2016.2616357

[24] Mead C.Neuromorphic electronic systems. Proc IEEE. 1990; 78(10):1629-1636. doi: 10.1109/5.58356

[25] Maass W. Networks of spiking neurons: The third generation of neural network models. Neural Netw. 1997; 10(9):1659-1671. doi: 10.1016/S0893-6080(97)00011-7

[26] Boahen KA. Point-to-point connectivity between neuromorphic chips using address events. IEEE Trans Circuits Syst II Analog Digit Signal Process. 2000; 47(5):416-434. doi: 10.1109/82.842110

[27] Diehl PU, Cook M. Unsupervised learning of digit recognition using spike-timing-dependent plasticity. Front Comput Neurosci. 2015; 9:99. doi: 10.3389/fncom.2015.00099

[28] Lapicque L. Recherches quantitatives sur l’excitation électrique des nerfs traitée comme une polarisation. J Physiol Pathol Gén. 1907; 9:620-634.

[29] McCulloch WS, Pitts W. A logical calculus of the ideas immanent in nervous activity. Bull Math Biophys. 1943; 5(4):115-133. doi: 10.1007/BF02478259

[30] Merolla PA, Arthur JV, Alvarez-Icaza R, et al. A million spiking-neuron integrated circuit with a scalable communication network and interface. Science. 2014; 345(6197):668-673. doi: 10.1126/science.1254642

[31] Davies M, Srinivasa N, Lin TH, et al. Loihi: A neuromorphic manycore processor with on-chip learning. IEEE Micro. 2018; 38(1):82-99. doi: 10.1109/MM.2018.112130359

[32] Strukov DB, Snider GS, Stewart DR, Williams RS.The missing memristor found. Nature. 2008; 453(7191):80-83. doi: 10.1038/nature06932

[33] Tuma T, Pantazi A, Le Gallo M, Sebastian A, Eleftheriou E.Stochastic phase-change neurons. Nat Nanotechnol. 2016; 11(8):693-699. doi: 10.1038/nnano.2016.70

[34] Kent AD, Worledge DC. A new spin on magnetic memories. Nat Nanotechnol. 2015; 10(3):187-191. doi: 10.1038/nnano.2015.24

[35] Locatelli N, Cros V, Grollier J.Spin-torque building blocks. Nat Mater. 2014; 13(1):11-20. doi: 10.1038/nmat3823

[36] Grollier J, Querlioz D, Camsari KY, Everschor-Sitte K, Fukami S, Stiles MD. Neuromorphic spintronics. Nat Electron. 2020; 3(7):360-370. doi: 10.1038/s41928-019-0360-9

[37] Torrejon J, Riou M, Araujo FA, et al. Neuromorphic computing with nanoscale spintronic oscillators. Nature. 2017; 547(7664):428-431. doi: 10.1038/nature23011

[38] Liu L, Pai CF, Li Y, Tseng HW, Ralph DC, Buhrman RA. Spin-torque switching with the giant spin Hall effect of tantalum. Science. 2012; 336(6081):555-558. doi: 10.1126/science.1218197

[39] Aadit NA, Grimaldi A, Carpentieri M, et al. Massively parallel probabilistic computing with sparse Ising machines. Nat Electron. 2022; 5(7):460-468. doi: 10.1038/s41928-022-00774-2

[40] Brunner D, Soriano MC, Mirasso CR, Fischer I. Parallel photonic information processing at gigabyte per second data rates using transient states. Nat Commun. 2013; 4:1364. doi: 10.1038/ncomms2368

[41] Chua LO. Memristor - the missing circuit element. IEEE Trans Circuit Theory. 1971; 18(5):507-519. doi: 10.1109/TCT.1971.1083337

[42] Soliman T, Chatterjee S, Laleni N, et al. First demonstration of in-memory computing crossbar using multi-level Cell FeFET. Nat Commun. 2023; 14:6348. doi: 10.1038/s41467-023-42110-y

[43] Feldmann J, Youngblood N, Wright CD, Bhaskaran H, Pernice WHP. All-optical spiking neurosynaptic networks with self-learning capabilities. Nature. 2019; 569(7755):208-214. doi: 10.1038/s41586-019-1157-8

[44] Le Gallo M, Nandakumar SR, Ciric L, et al. Precision of bit slicing with in-memory computing based on analog phase-change memory crossbars. Neuromorph Comput Eng. 2022; 2(1):014009. doi: 10.1088/2634-4386/ac4fb7

[45] Gerstner W, Kistler WM, Naud R, Paninski L. Neuronal Dynamics: From Single Neurons to Networks and Models of Cognition. Cambridge University Press; 2014. doi: 10.1017/CBO9781107447615

[46] Indiveri G, Linares-Barranco B, Hamilton TJ, et al. Neuromorphic silicon neuron circuits. Front Neurosci. 2011; 5:73. doi: 10.3389/fnins.2011.00073

[47] Csaba G, Porod W. Coupled oscillators for computing: A review and perspective. Appl Phys Rev. 2020; 7(1):011302. doi: 10.1063/1.5120412

[48] Xu X, Tan M, Corcoran B, et al. 11 TOPS photonic convolutional accelerator for optical neural networks. Nature. 2021; 589(7840):44-51. doi: 10.1038/s41586-020-03063-0

[49] Borders WA, Pervaiz AZ, Fukami S, Camsari KY, Ohno H, Datta S. Integer factorization using stochastic magnetic tunnel junctions. Nature. 2019; 573(7774):390-393. doi: 10.1038/s41586-019-1557-9

[50] Shastri BJ, Tait AN, Ferreira de Lima T, et al. Photonics for artificial intelligence and neuromorphic computing. Nat Photonics. 2021; 15(2):102-114. doi: 10.1038/s41566-020-00754-y

[51] Le Gallo M, Khaddam-Aljameh R, Stanisavljevic M, et al. A 64-core mixed-signal in-memory compute chip based on phase-change memory for deep neural network inference. Nat Electron. 2023; 6(9):680-693. doi: 10.1038/s41928-023-01010-1

[52] Hopfield JJ. Neural networks and physical systems with emergent collective computational abilities. Proc Natl Acad Sci USA. 1982; 79(8):2554-2558. doi: 10.1073/pnas.79.8.2554

[53] Esser SK, Merolla PA, Arthur JV, et al. Convolutional networks for fast, energy-efficient neuromorphic computing. Proc Natl Acad Sci USA. 2016; 113(41):11441-11446. doi: 10.1073/pnas.1604850113

[54] Wang T, Wu L, Nobel P, Roychowdhury J. Solving combinatorial optimisation problems using oscillator based Ising machines. Nat Comput. 2021; 20(2):287-306. doi: 10.1007/s11047-021-09845-3

[55] Zenke F, Ganguli S. SuperSpike: Supervised learning in multilayer spiking neural networks. Neural Comput. 2018; 30(6):1514-1541. doi: 10.1162/neco_a_01086

[56] Le Gallo M, Sebastian A, Mathis R, et al. Mixed-precision in-memory computing. Nat Electron. 2018; 1:246-253. doi: 10.1038/s41928-018-0054-8

[57] Suri M, Bichler O, Querlioz D, et al. Bio-inspired stochastic computing using binary CBRAM synapses. IEEE Trans Electron Devices. 2013; 60(7):2402-2409. doi: 10.1109/TED.2013.2263000

[58] Neftci EO, Mostafa H, Zenke F. Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based optimization to spiking neural networks. IEEE Signal Process Mag. 2019; 36(6):51-63. doi: 10.1109/MSP.2019.2931595

[59] Jacob B, Kligys S, Chen B, et al. Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE; 2018:2704-2713. doi: 10.1109/cvpr.2018.00286

[60] Bohte SM, Kok JN, La Poutré H. Error-backpropagation in temporally encoded networks of spiking neurons. Neurocomputing. 2002; 48(1-4):17-37. doi: 10.1016/S0925-2312(01)00658-0

[61] Alibart F, Zamanidoost E, Strukov DB. Pattern classification by memristive crossbar circuits using ex situ and in situ training. Nat Commun. 2013; 4:2072. doi: 10.1038/ncomms3072

[62] Gupta S, Agrawal A, Gopalakrishnan K, Narayanan P. Deep learning with limited numerical precision. In: International Conference on Machine Learning. 2015:1737-1746.

[63] Hinton GE, Sejnowski TJ. Learning and relearning in Boltzmann machines. In: Rumelhart DE, McClelland JL eds. Parallel Distributed Processing, Vol 1. MIT Press; 1986:282-317.