The deployment of large-scale 32-bit neural networks is hindered by the escalating parameter sizes and inherent susceptibility to noise. Here, we present a non-volatile memory (NVM)-based probabilistic computing architecture, which harnesses both the stochastic nature and in-memory computing capabilities of NVM to enhance computation efficiency and robustness. Utilizing in-plane magnetized spin-orbit torque magneto-resistive random-access memory (i-SOT MRAM) cells, we demonstrate ultra-fast (400 ps) all-electrical switching, 1011 data endurance, and voltage-controllable probabilistic states with a low variation of 3.8%. These features enable the implementation of an on-chip SOT probabilistic binary neural network (SOT-PBNN) hardware, achieving near-baseline accuracy (88.2%) on CIFAR-10 classification tasks. Moreover, the stochastic training process endows the SOT-PBNN to attain a 7-fold improvement in classification accuracy over 32-bit CNN counterpart under a 25% write/read noise, while reducing the parameter size and inference energy by 1-2 orders. Our work establishes a lightweight framework suitable for realizing artificial intelligence platforms on resource-constrained edge devices.
Hongchao Zhang, Tianxiao Nie, Weisheng Zhao et al.· AIP Publishing· 0 citations
The deployment of large-scale 32-bit neural networks is hindered by the escalating parameter sizes and inherent susceptibility to noise. Here, we present a non-volatile memory (NVM)-based probabilistic computing architecture, which harnesses both the stochastic nature and in-memory computing capabilities of NVM to enhance computation efficiency and robustness. Utilizing in-plane magnetized spin-orbit torque magneto-resistive random-access memory (i-SOT MRAM) cells, we demonstrate ultra-fast (400 ps) all-electrical switching, 1011 data endurance, and voltage-controllable probabilistic states with a low variation of 3.8%. These features enable the implementation of an on-chip SOT probabilistic binary neural network (SOT-PBNN) hardware, achieving near-baseline accuracy (88.2%) on CIFAR-10 classification tasks. Moreover, the stochastic training process endows the SOT-PBNN to attain a 7-fold improvement in classification accuracy over 32-bit CNN counterpart under a 25% write/read noise, while reducing the parameter size and inference energy by 1-2 orders. Our work establishes a lightweight framework suitable for realizing artificial intelligence platforms on resource-constrained edge devices.
Hongchao Zhang, Tianxiao Nie, Weisheng Zhao et al.· AIP Publishing· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
