Plasticity-Enhanced Domain-Wall MTJ Neural Networks for Energy-Efficient Online Learning

Machine learning implements backpropagation via abundant training samples. We demonstrate a multi-stage learning system realized by a promising non-volatile memory device, the domain-wall magnetic tunnel junction (DW-MTJ). The system consists of unsupervised (clustering) as well as supervised sub-systems, and generalizes quickly (with few samples). We demonstrate interactions between physical properties of this device and optimal implementation of neuroscience-inspired plasticity learning rules, and highlight performance on a suite of tasks. Our energy analysis confirms the value of the approach, as the learning budget stays below 20 $\mu J$ even for large tasks used typically in machine learning.

[1]  Jacques-Olivier Klein,et al.  Spin-Transfer Torque Magnetic Memory as a Stochastic Memristive Synapse for Neuromorphic Systems , 2015, IEEE Transactions on Biomedical Circuits and Systems.

[2]  Roland Vollgraf,et al.  Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms , 2017, ArXiv.

[3]  Saima Afroz Siddiqui,et al.  The Spatial Resolution Limit for an Individual Domain Wall in Magnetic Nanowires. , 2017, Nano letters.

[4]  Pritish Narayanan,et al.  Neuromorphic computing using non-volatile memory , 2017 .

[5]  Stephen B. Furber,et al.  Biologically Inspired Means for Rank-Order Encoding Images: A Quantitative Analysis , 2010, IEEE Transactions on Neural Networks.

[6]  Joseph S. Friedman,et al.  Three-terminal magnetic tunnel junction synapse circuits showing spike-timing-dependent plasticity , 2019, Journal of Physics D: Applied Physics.

[7]  Jennifer Hasler,et al.  Vector-Matrix Multiply and Winner-Take-All as an Analog Classifier , 2014, IEEE Transactions on Very Large Scale Integration (VLSI) Systems.

[8]  Mark Horowitz,et al.  1.1 Computing's energy problem (and what we can do about it) , 2014, 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC).

[9]  Manan Suri,et al.  Design Exploration of IoT centric Neural Inference Accelerators , 2018, ACM Great Lakes Symposium on VLSI.

[10]  Wolfgang Maass,et al.  On the Computational Power of Winner-Take-All , 2000, Neural Computation.

[11]  Rajesh P. N. Rao,et al.  Spike-Timing-Dependent Hebbian Plasticity as Temporal Difference Learning , 2001, Neural Computation.

[12]  André van Schaik,et al.  An Online Learning Algorithm for Neuromorphic Hardware Implementation , 2015, ArXiv.

[13]  Wolfgang Maass,et al.  Neural Computation with Winner-Take-All as the Only Nonlinear Operation , 1999, NIPS.

[14]  Steven J. Plimpton,et al.  Multiscale Co-Design Analysis of Energy, Latency, Area, and Accuracy of a ReRAM Analog Neural Training Accelerator , 2017, IEEE Journal on Emerging and Selected Topics in Circuits and Systems.

[15]  Kyeong-Sik Min,et al.  Ta2O5-memristor synaptic array with winner-take-all method for neuromorphic pattern matching , 2016 .

[16]  Joseph S. Friedman,et al.  Magnetic domain wall neuron with lateral inhibition , 2018, Journal of Applied Physics.

[17]  Zhigang Zeng,et al.  Analysis and design of winner-take-all behavior based on a novel memristive neural network , 2013, Neural Computing and Applications.

[18]  C. Ross,et al.  Low Energy Magnetic Domain Wall Logic in Short, Narrow, Ferromagnetic Wires , 2012, IEEE Magnetics Letters.

[19]  Damien Querlioz,et al.  Bioinspired Programming of Memory Devices for Implementing an Inference Engine , 2015, Proceedings of the IEEE.

[20]  Davide Anguita,et al.  A Public Domain Dataset for Human Activity Recognition using Smartphones , 2013, ESANN.

[21]  David Kappel,et al.  STDP Installs in Winner-Take-All Circuits an Online Approximation to Hidden Markov Model Learning , 2014, PLoS Comput. Biol..

[22]  Jan Vogel,et al.  Tuning domain wall velocity with Dzyaloshinskii-Moriya interaction , 2017, 1709.09857.

[23]  Alvaro Velasquez,et al.  Unsupervised Competitive Hardware Learning Rule for Spintronic Clustering Architecture , 2020, ArXiv.

[24]  John J. Hopfield,et al.  Unsupervised learning by competing hidden units , 2018, Proceedings of the National Academy of Sciences.

[25]  Upasana Sahu,et al.  Comparing domain wall synapse with other Non Volatile Memory devices for on-chip learning in Analog Hardware Neural Network , 2019, AIP Advances.

[26]  R. Engel-Herbert,et al.  Calculation of the magnetic stray field of a uniaxial magnetic domain , 2005 .

[27]  Farnood Merrikh-Bayat,et al.  Efficient training algorithms for neural networks based on memristive crossbar circuits , 2015, 2015 International Joint Conference on Neural Networks (IJCNN).

[28]  Yves Chauvin,et al.  Backpropagation: the basic theory , 1995 .

[29]  Christopher H. Bennett,et al.  Maximized lateral inhibition in paired magnetic domain wall racetracks for neuromorphic computing , 2019, Nanotechnology.

[30]  Stephen Grossberg,et al.  Competitive Learning: From Interactive Activation to Adaptive Resonance , 1987, Cogn. Sci..

[31]  Dongsuk Jeon,et al.  7.6 A 65nm 236.5nJ/Classification Neuromorphic Processor with 7.5% Energy Overhead On-Chip Learning Using Direct Spike-Only Feedback , 2019, 2019 IEEE International Solid- State Circuits Conference - (ISSCC).

[32]  Damien Querlioz,et al.  Contrasting Advantages of Learning With Random Weights and Backpropagation in Non-Volatile Memory Neural Networks , 2019, IEEE Access.

[33]  Joseph S. Friedman,et al.  Semi-supervised learning and inference in domain-wall magnetic tunnel junction (DW-MTJ) neural networks , 2019, NanoScience + Engineering.

[34]  Jacques-Olivier Klein,et al.  Bioinspired networks with nanoscale memristive devices that combine the unsupervised and supervised learning approaches , 2012, 2012 IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH).