Resistive Crossbars as Approximate Hardware Building Blocks for Machine Learning: Opportunities and Challenges
The DOI matches its Crossref record; 173 of the 178 references it cites were found in a registry or an index.
Abstract
Traditional computing systems based on the von Neumann architecture are fundamentally bottlenecked by data transfers between processors and memory. The emergence of data-intensive workloads, such as machine learning (ML), creates an urgent need to address this bottleneck by designing computing platforms that utilize the principle of colocated memory and processing units. Such an approach, known as “in-memory computing,” can potentially eliminate data movement costs by computing inside the memory array itself. Crossbars based on resistive nonvolatile memory (NVM) devices have shown immense promise in serving as the building blocks of in-memory computing systems for ML workloads. This is because their high density can lead to higher on-chip storage capacity, while they can also perform massively parallel, in situ matrix-vector multiplication (MVM) operations, thereby accelerating the main computational kernel of ML workloads. However, resistive crossbar-based analog computing is inherently approximate due to the device- and circuit-level nonidealities. Furthermore, the area and energy costs of peripheral circuits for conversions between the analog and digital domains can greatly diminish the intrinsic efficiency of crossbar-based MVM computation. We present a comprehensive overview of the emerging paradigm of computing using NVM crossbars for accelerating ML workloads. We describe the design principles of resistive crossbars, including the devices and associated circuits that constitute them. We discuss intrinsic approximations arising from the device and circuit characteristics and study their functional impact on the MVM operation. Next, we present an overview of spatial architectures that exploit the high storage density of NVM crossbars. Furthermore, we elaborate on software frameworks that effectively capture device-circuit-architecture characteristics to evaluate the performance of large-scale deep neural networks (DNNs) using resistive crossbar-based hardware. Finally, we discuss open challenges and future research directions that need to be explored in order to realize the vision of resistive crossbars as the building blocks of future computing platforms.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“However, the disadvantage of time-encoding schemes is the reduction of computation speed and hardware overhead required for partial sums computation96.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [96]
In the citing paper
“The resolution of ADC required to represent all possible outputs of the VMM operation depends on input precision (DAC resolution), number of crossbar rows , and precision of the weights cells (conductance resolution), and can be calculated as 96.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [96]
In the citing paper
“ADC sharing requires additional multiplexers and sample-and-hold circuits and also increases latency96 (i.e. more time is required to process each input pattern, thus reducing the throughput of the ANN).”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [96]
In the citing paper
“In binarized networks, ADC can be replaced by a 1-bit comparator96 or ADC-like multi-level sense amplifier158.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [96]
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“Deep neural networks (DNNs), in particular, have had enormous success in several ML tasks, such as image classification [2], [3], object detection [4], and natural language processing [5].”
BackgroundImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 2×
“[2] A.”
BackgroundPyTorch: An Imperative Style, High-Performance Deep Learning Library
Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al. (2019) · arXiv · doi:10.48550/arxiv.1912.01703 · Cited by 31,473 · cited 1×
“One approach to perform this is functionally abstract the nonideal MVM operation and integrate it into existing ML frameworks (such as PyTorch [127], Tensor-flow [128], and Caffe [129]).”
MethodGirshick, R. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.169 · Cited by 28,105
TensorFlow: A system for large-scale machine learning
Abadi, M., Barham, P., Chen, J., Chen, Z., Davis, A., Dean, J., Devin, M., Ghemawat, S., Irving, G., Isard, M., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1605.08695 · Cited by 11,822
Caffe: Convolutional Architecture for Fast Feature Embedding
Jia, Y., Shelhamer, E., Donahue, J., Karayev, S., Long, J., Girshick, R., Guadarrama, S., & Darrell, T. (2014) · Proceedings of the 22nd ACM international conference on Multimedia · doi:10.1145/2647868.2654889 · Cited by 11,077 · cited 1×
“One approach to perform this is functionally abstract the nonideal MVM operation and integrate it into existing ML frameworks (such as PyTorch [127], Tensor-flow [128], and Caffe [129]).”
MethodMastering the game of Go without human knowledge
Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I., Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M., Bolton, A., et al. (2017) · Nature · doi:10.1038/nature24270 · Cited by 9,084 · cited 1×
“Rapid advances in machine learning (ML) and artificial intelligence (AI) have touched human lives in an unprecedented way as the application space has proliferated over the last decade [1].”
BackgroundAnisotropic Superexchange Interaction and Weak Ferromagnetism
Moriya, T. (1960) · Physical Review · doi:10.1103/physrev.120.91 · Cited by 7,234 · cited 1×
“While the stability of the Neel wall is aided by the Dzyaloshinskii–Moriya interaction [67], [68], the Bloch walls need external magnetic fields for proper operation [69].”
MethodCurrent-driven excitation of magnetic multilayers
Slonczewski, J. C. (1996) · Journal of Magnetism and Magnetic Materials · doi:10.1016/0304-8853(96)00062-5 · Cited by 6,825
A thermodynamic theory of “weak” ferromagnetism of antiferromagnetics
Dzyaloshinsky, I. (1958) · Journal of Physics and Chemistry of Solids · doi:10.1016/0022-3697(58)90076-3 · Cited by 6,259 · cited 1×
“While the stability of the Neel wall is aided by the Dzyaloshinskii–Moriya interaction [67], [68], the Bloch walls need external magnetic fields for proper operation [69].”
MethodIn-Datacenter Performance Analysis of a Tensor Processing Unit
Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., et al. (2017) · Proceedings of the 44th Annual International Symposium on Computer Architecture · doi:10.1145/3079856.3080246 · Cited by 4,537 · cited 1×
“Designers have proposed special-purpose accelerators [7]–[11] to improve the time and energy required to execute DNNs.”
BackgroundXNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks
Rastegari, M., Ordonez, V., Redmon, J., & Farhadi, A. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_32 · Cited by 4,408 · cited 2×
“As a result, researchers have explored CMOS macros primarily in the context of binary neural networks [107], [108] where the matrix–vector multiplications are represented using XNOR operations [57], [109] between binary inputs and weights.”
BackgroundNanoscale Memristor Device as Synapse in Neuromorphic Systems
Jo, S. H., Chang, T., Ebong, I., Bhadviya, B. B., Mazumder, P., & Lu, W. (2010) · Nano Letters · doi:10.1021/nl904092h · Cited by 4,174 · cited 1×
“Typical CBRAMs are based on Ag electrodes where the multilevel behavior is achieved by tuning the volume of Ag-rich and Ag-poor regions [55].”
BackgroundSpin-Torque Switching with the Giant Spin Hall Effect of Tantalum
Liu, L., Pai, C.-F., Li, Y., Tseng, H. W., Ralph, D. C., & Buhrman, R. A. (2012) · Science · doi:10.1126/science.1218197 · Cited by 4,108 · cited 1×
“By stacking an MTJ on top of the heavy metal (HM) layer, one can write into the FM layer by passing a current through the HM using a phenomenon called the spin Hall effect (SHE) [65].”
MethodEyeriss: An Energy-Efficient Reconfigurable Accelerator for Deep Convolutional Neural Networks
Chen, Y.-H., Krishna, T., Emer, J. S., & Sze, V. (2017) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2016.2616357 · Cited by 3,247 · cited 1×
“[9] Y.”
Wong, H.-S. P., Lee, H.-Y., Yu, S., Chen, Y.-S., Wu, Y., Chen, P.-S., Lee, B., Chen, F. T., & Tsai, M.-J. (2012) · Proceedings of the IEEE · doi:10.1109/jproc.2012.2190369 · Cited by 2,757 · cited 2×
“[38] H.”
In-memory computing with resistive switching devices
Ielmini, D., & Wong, H.-S. P. (2018) · Nature Electronics · doi:10.1038/s41928-018-0092-2 · Cited by 2,298 · cited 1×
“This led to a growing interest in exploring NVM technology as the substrate for the next generation of ML hardware [19], [25], [26].”
BackgroundFully hardware-implemented memristor convolutional neural network
Yao, P., Wu, H., Gao, B., Tang, J., Zhang, Q., Zhang, W., Yang, J. J., & Qian, H. (2020) · Nature · doi:10.1038/s41586-020-1942-4 · Cited by 2,179 · cited 1×
“[97] P.”
The future of electronics based on memristive systems
Zidan, M. A., Strachan, J. P., & Lu, W. D. (2018) · Nature Electronics · doi:10.1038/s41928-017-0006-8 · Cited by 2,138 · cited 1×
“This led to a growing interest in exploring NVM technology as the substrate for the next generation of ML hardware [19], [25], [26].”
BackgroundCourbariaux, M., Hubara, I., Soudry, D., El-Yaniv, R., & Bengio, Y. (2016) · arXiv · doi:10.48550/arxiv.1602.02830 · Cited by 2,106
Wong, H.-S. P., Raoux, S., Kim, S., Liang, J., Reifenberg, J. P., Rajendran, B., Asheghi, M., & Goodson, K. E. (2010) · Proceedings of the IEEE · doi:10.1109/jproc.2010.2070050 · Cited by 1,806
Hitting the memory wall: implications of the obvious
Wulf, W. A., & McKee, S. A. (1995) · ACM SIGARCH Computer Architecture News · doi:10.1145/216585.216588 · Cited by 1,624 · cited 1×
“Despite their extensive use of on-chip scratchpads to exploit data reuse, these accelerators are eventually still limited by the off-chip memory bottleneck [12], motivating closer integration of memory and processing units as the next step in the quest for further efficiency.”
BackgroundIkeda, S., Hayakawa, J., Ashizawa, Y., Lee, Y. M., Miura, K., Hasegawa, H., Tsunoda, M., Matsukura, F., & Ohno, H. (2008) · Applied Physics Letters · doi:10.1063/1.2976435 · Cited by 1,596 · cited 2×
“[46] S.”
MethodISAAC: a convolutional neural network accelerator with in-situ analog arithmetic in crossbars
Shafiee, A., Nag, A., Muralimanohar, N., Balasubramonian, R., Strachan, J. P., Hu, M., Williams, R. S., & Srikumar, V. (2016) · ACM SIGARCH Computer Architecture News · doi:10.1145/3007787.3001139 · Cited by 1,536 · cited 8×
“For instance, ISAAC [29] and PUMA [30] adopted an 8-bit SAR ADC for higher precision computations, while XNOR-RRAM implemented a 3-bit flash ADC for binary input/weight computation [57].”
BackgroundMethodInfluentialDocument Modeling with Gated Recurrent Neural Network for Sentiment Classification
Tang, D., Qin, B., & Liu, T. (2015) · Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d15-1167 · Cited by 1,519 · cited 2×
“[5] D.”
BackgroundHitting the Memory Wall: Implications of the Obvious
Wulf, W. A., & McKee, S. A. (1995) · Cited by 1,379
Theory of tunneling magnetoresistance of an epitaxial Fe/MgO/Fe(001) junction
Mathon, J., & Umerski, A. (2001) · Physical Review B · doi:10.1103/physrevb.63.220403 · Cited by 1,302 · cited 1×
“Although TMR of 600% has been achieved experimentally for MTJs [46] and theoretical predicting that 1000% can be achievable [62], [63], large-scale demonstrations are still lacking.”
BackgroundAnalogue signal and image processing with large memristor crossbars
Li, C., Hu, M., Li, Y., Jiang, H., Ge, N., Montgomery, E., Zhang, J., Song, W., Dávila, N., Graves, C. E., et al. (2017) · Nature Electronics · doi:10.1038/s41928-017-0002-z · Cited by 1,292 · cited 2×
“Like PCMs, RRAMs have high density and have been successfully demonstrated to achieve crossbar sizes up to 128 × 128 [56].”
BackgroundChi, P., Li, S., Xu, C., Zhang, T., Zhao, J., Liu, Y., Wang, Y., & Xie, Y. (2016) · ACM SIGARCH Computer Architecture News · doi:10.1145/3007787.3001140 · Cited by 1,231 · cited 2×
“Typical ReRam-based accelerators are spatial and organized in hierarchies: core, multicore, and node [29], [30], [136].”
Micikevicius, P., Narang, S., Alben, J., Diamos, G., Elsen, E., Garcia, D., Ginsburg, B., Houston, M., Kuchaiev, O., Venkatesh, G., et al. (2017) · arXiv · doi:10.48550/arxiv.1710.03740 · Cited by 1,179 · cited 1×
“Note that for larger data sets, such as ImageNet, precision requirements can be much higher [150].”
BackgroundNeuro-Inspired Computing With Emerging Nonvolatile Memorys
Yu, S. (2018) · Proceedings of the IEEE · doi:10.1109/jproc.2018.2790840 · Cited by 1,170 · cited 2×
“This led to a growing interest in exploring NVM technology as the substrate for the next generation of ML hardware [19], [25], [26].”
BackgroundEquivalent-accuracy accelerated neural-network training using analogue memory
Ambrogio, S., Narayanan, P., Tsai, H., Shelby, R. M., Boybat, I., di Nolfo, C., Sidler, S., Giordano, M., Bodini, M., Farinha, N. C. P., et al. (2018) · Nature · doi:10.1038/s41586-018-0180-5 · Cited by 1,151 · cited 3×
“In fact, researchers have been successful in experimentally demonstrating large-scale PCM crossbars for ML applications [21], [43].”
BackgroundBurr, G. W., Shelby, R. M., Sidler, S., di Nolfo, C., Jang, J., Boybat, I., Shenoy, R. S., Narayanan, P., Virwani, K., Giacometti, E. U., et al. (2015) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2015.2439635 · Cited by 936 · cited 2×
“In fact, researchers have been successful in experimentally demonstrating large-scale PCM crossbars for ML applications [21], [43].”
BackgroundParallel programming of an ionic floating-gate memory array for scalable neuromorphic computing
Fuller, E. J., Keene, S. T., Melianas, A., Wang, Z., Agarwal, S., Li, Y., Tuchman, Y., James, C. D., Marinella, M. J., Yang, J. J., et al. (2019) · Science · doi:10.1126/science.aaw5581 · Cited by 756 · cited 2×
“Such a scheme has also been experimentally demonstrated on floating gate memory arrays [105].”
BackgroundMemristor‐Based Analog Computation and Neural Network Classification with a Dot Product Engine
Hu, M., Graves, C. E., Li, C., Li, Y., Ge, N., Montgomery, E., Davila, N., Jiang, H., Williams, R. S., Yang, J. J., et al. (2018) · Advanced Materials · doi:10.1002/adma.201705914 · Cited by 751 · cited 2×
“[22] M.”
BackgroundA fully integrated reprogrammable memristor–CMOS system for efficient multiply–accumulate operations
Cai, F., Correll, J. M., Lee, S. H., Lim, Y., Bothra, V., Zhang, Z., Flynn, M. P., & Lu, W. D. (2019) · Nature Electronics · doi:10.1038/s41928-019-0270-x · Cited by 728 · cited 2×
“[23] F.”
BackgroundHu, M., Strachan, J. P., Li, Z., Grafals, E. M., Davila, N., Graves, C., Lam, S., Ge, N., Yang, J. J., & Williams, R. S. (2016) · Proceedings of the 53rd Annual Design Automation Conference · doi:10.1145/2897937.2898010 · Cited by 686 · cited 6×
“On the other hand, one-transistor–one-memristive (1T-1R) device crossbars have exhibited MVM computations with fewer errors [88], [89], in comparison to 1S-1R crossbars.”
BackgroundMethodInfluentialChen, P.-Y., Peng, X., & Yu, S. (2018) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2018.2789723 · Cited by 652 · cited 1×
“[148].”
Ielmini, D., & Zhang, Y. (2007) · Journal of Applied Physics · doi:10.1063/1.2773688 · Cited by 642 · cited 1×
“For example, transport in amorphous PCM devices has been described by the Poole–Frenkel transport [111] of carriers, which results in the current being linear at small voltages and exponential at higher voltages.”
BackgroundA FeFET based super-low-power ultra-fast embedded NVM technology for 22nm FDSOI and beyond
Dunkel, S., Trentzsch, M., Richter, R., Moll, P., Fuchs, C., Gehring, O., Majer, M., Wittek, S., Muller, B., Melde, T., et al. (2017) · 2017 IEEE International Electron Devices Meeting (IEDM) · doi:10.1109/iedm.2017.8268425 · Cited by 540 · cited 1×
“Even though FE materials have been explored as a nonvolatile capacitor for a long time [71], recent devices propose integration of FE layer in the gate-stack of a standard transistor [47], [72], [73].”
BackgroundIn-Memory Computation of a Machine-Learning Classifier in a Standard 6T SRAM Array
Zhang, J., Wang, Z., & Verma, N. (2017) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2016.2642198 · Cited by 532 · cited 1×
“[17] J.”
Watanabe, Y., Bednorz, J. G., Bietsch, A., Gerber, C., Widmer, D., Beck, A., & Wind, S. J. (2001) · Applied Physics Letters · doi:10.1063/1.1377617 · Cited by 522 · cited 2×
“[48] Y.”
BackgroundMoons, B., Uytterhoeven, R., Dehaene, W., & Verhelst, M. (2017) · 2017 IEEE International Solid-State Circuits Conference (ISSCC) · doi:10.1109/isscc.2017.7870353 · Cited by 479
Chen, Z., & Liu, B. (2018) · Synthesis Lectures on Artificial Intelligence and Machine Learning · doi:10.1007/978-3-031-01581-6 · Cited by 466
Scaling for edge inference of deep neural networks
Xu, X., Ding, Y., Hu, S. X., Niemier, M., Cong, J., Hu, Y., & Shi, Y. (2018) · Nature Electronics · doi:10.1038/s41928-018-0059-3 · Cited by 462 · cited 1×
“However, the capabilities of deep learning algorithms have been accompanied by an increase in computational complexity [6].”
BackgroundXNOR-SRAM: In-Memory Computing SRAM Macro for Binary/Ternary Deep Neural Networks
Yin, S., Jiang, Z., Seo, J.-S., & Seok, M. (2020) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2019.2963616 · Cited by 429
PUMA: A Programmable Ultra-efficient Memristor-based Accelerator for Machine Learning Inference
Ankit, A., Hajj, I. E., Chalamalasetti, S. R., Ndu, G., Foltin, M., Williams, R. S., Faraboschi, P., Hwu, W.-M. W., Strachan, J. P., Roy, K., et al. (2019) · Proceedings of the Twenty-Fourth International Conference on Architectural Support for Programming Languages and Operating Systems · doi:10.1145/3297858.3304049 · Cited by 424 · cited 10×
“This promise has fueled great interest over the past decade in designing systems that can leverage the high computing efficiency as well as the high-density storage of crossbar-based MVM units [28]–[30].”
BackgroundMethodInfluentialSelf-citationDevice modeling of ferroelectric capacitors
Miller, S. L., Nasby, R. D., Schwank, J. R., Rodgers, M. S., & Dressendorfer, P. V. (1990) · Journal of Applied Physics · doi:10.1063/1.346845 · Cited by 379 · cited 1×
“Even though FE materials have been explored as a nonvolatile capacitor for a long time [71], recent devices propose integration of FE layer in the gate-stack of a standard transistor [47], [72], [73].”
BackgroundAnatomy of Ag/Hafnia‐Based Selectors with 1010 Nonlinearity
Midya, R., Wang, Z., Zhang, J., Savel'ev, S. E., Li, C., Rao, M., Jang, M. H., Joshi, S., Jiang, H., Lin, P., et al. (2017) · Advanced Materials · doi:10.1002/adma.201604457 · Cited by 379 · cited 1×
“[83] R.”
Low-field amorphous state resistance and threshold voltage drift in chalcogenide materials
Pirovano, A., Lacaita, A. L., Pellizzer, F., Kostylev, S. A., Benvenuti, A., & Bez, R. (2004) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2004.825805 · Cited by 377 · cited 2×
“Moreover, resistance drifting phenomenon [162] in PCM technology can affect data retention, which can have a detrimental effect on NVM crossbars when used as MVM units.”
MethodHighly reliable TaOx ReRAM and direct evidence of redox reaction mechanism
Wei, Z., Kanzawa, Y., Arita, K., Katoh, Y., Kawai, K., Muraoka, S., Mitani, S., Fujii, S., Katayama, K., Iijima, M., et al. (2008) · 2008 IEEE International Electron Devices Meeting · doi:10.1109/iedm.2008.4796676 · Cited by 337 · cited 1×
“…to be addressed such as resistance drifting due to structural relaxations after the melt-quench amorphization. b) Metal–oxide materials and CBRAMs: Perovskite oxides, such as SrTiO 3 , and binary metal oxides, such as HfO x , TiO x , and TaO x , exhibit resistive switching properties [48]–[51].”
BackgroundBiswas, A., & Chandrakasan, A. P. (2018) · 2018 IEEE International Solid - State Circuits Conference - (ISSCC) · doi:10.1109/isscc.2018.8310397 · Cited by 337 · cited 4×
“2) CMOS MVM Macros: It is worth mentioning that MVM units can also be constructed using CMOS-based SRAM cells [13]–[15], [106].”
BackgroundInfluentialIdentification of a determining parameter for resistive switching of TiO2 thin films
Rohde, C., Choi, B. J., Jeong, D. S., Choi, S., Zhao, J.-S., & Hwang, C. S. (2005) · Applied Physics Letters · doi:10.1063/1.1968416 · Cited by 333 · cited 1×
“[50] C.”
Serving DNNs in Real Time at Datacenter Scale with Project Brainwave
Chung, E., Fowers, J., Ovtcharov, K., Papamichael, M., Caulfield, A., Massengill, T., Liu, M., Lo, D., Alkalay, S., Haselman, M., et al. (2018) · IEEE Micro · doi:10.1109/mm.2018.022071131 · Cited by 328 · cited 1×
“[8] E.”
Liu, Q., Gao, B., Yao, P., Wu, D., Chen, J., Pang, Y., Zhang, W., Liao, Y., Xue, C.-X., Chen, W.-H., et al. (2020) · 2020 IEEE International Solid- State Circuits Conference - (ISSCC) · doi:10.1109/isscc19947.2020.9062953 · Cited by 308 · cited 1×
“[98] Q.”
Lee, J., Kim, C., Kang, S., Shin, D., Kim, S., & Yoo, H.-J. (2018) · 2018 IEEE International Solid - State Circuits Conference - (ISSCC) · doi:10.1109/isscc.2018.8310262 · Cited by 303 · cited 1×
“[10] J.”
A SPICE Compact Model of Metal Oxide Resistive Switching Memory With Variations
Guan, X., Yu, S., & Wong, H.-S. P. (2012) · IEEE Electron Device Letters · doi:10.1109/led.2012.2210856 · Cited by 282 · cited 2×
“RRAM devices also exhibit nonlinear characteristics [59] that introduce errors in the MVM computations. c) Spintronic devices: Unlike PCMs and RRAMs, spin-tronic devices use electron spin as the storage variable.”
BackgroundIncentivizing Exploration In Reinforcement Learning With Deep Predictive Models
Stadie, B. C., Levine, S., & Abbeel, P. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1507.00814 · Cited by 280 · cited 1×
“Training based on small batch sizes is common in several emerging applications, such as lifelong learning [143] and online reinforcement learning [144], where training does not rely on any earlier collected data set.”
BackgroundNovel ferroelectric FET based synapse for neuromorphic systems
Mulaosmanovic, H., Ocker, J., Muller, S., Noack, M., Muller, J., Polakowski, P., Mikolajick, T., & Slesazeck, S. (2017) · 2017 Symposium on VLSI Technology · doi:10.23919/vlsit.2017.7998165 · Cited by 276 · cited 1×
“As a result, FeFETs have shown immense promise as neuromimetic devices [74]–[77].”
BackgroundRecovery and Drift Dynamics of Resistance and Threshold Voltages in Phase-Change Memories
Ielmini, D., Lacaita, A. L., & Mantegazza, D. (2007) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2006.888752 · Cited by 268 · cited 1×
“The origin of these variations has been studied in detailed for various technologies [163], [164].”
BackgroundRescuing Memristor-based Neuromorphic Design with High Defects
Liu, C., Hu, M., Strachan, J. P., & Li, H. (. (2017) · Proceedings of the 54th Annual Design Automation Conference 2017 · doi:10.1145/3061639.3062310 · Cited by 257 · cited 3×
“…can be of three types: 1) retraining of weights for error compensation [33], [34]; 2) mapping algorithms to reduce the impact of nonidealities [153]–[156]; and 3) hardware compensation [33]. a) Compensation through retraining: When large-scale DNNs are mapped to nonideal crossbars using the…”
BackgroundMethod24.5 A Twin-8T SRAM Computation-In-Memory Macro for Multiple-Bit CNN-Based Machine Learning
Si, X., Chen, J.-J., Tu, Y.-N., Huang, W.-H., Wang, J.-H., Chiu, Y.-C., Wei, W.-C., Wu, S.-Y., Sun, X., Liu, R., et al. (2019) · 2019 IEEE International Solid- State Circuits Conference - (ISSCC) · doi:10.1109/isscc.2019.8662392 · Cited by 254 · cited 2×
“2) CMOS MVM Macros: It is worth mentioning that MVM units can also be constructed using CMOS-based SRAM cells [13]–[15], [106].”
BackgroundWalczyk, C., Walczyk, D., Schroeder, T., Bertaud, T., Sowinska, M., Lukosius, M., Fraschke, M., Wolansky, D., Tillack, B., Miranda, E., et al. (2011) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2011.2160265 · Cited by 246 · cited 2×
“[126] C.”
BackgroundXNOR-RRAM: A scalable and parallel resistive synaptic architecture for binary neural networks
Sun, X., Yin, S., Peng, X., Liu, R., Seo, J.-S., & Yu, S. (2018) · 2018 Design, Automation & Test in Europe Conference & Exhibition (DATE) · doi:10.23919/date.2018.8342235 · Cited by 246 · cited 6×
“For instance, ISAAC [29] and PUMA [30] adopted an 8-bit SAR ADC for higher precision computations, while XNOR -RRAM implemented a 3-bit flash ADC for binary input/weight computation [57].”
BackgroundMethodInfluentialAccelerator-friendly neural-network training: Learning variations and defects in RRAM crossbar
Chen, L., Li, J., Chen, Y., Deng, Q., Shen, J., Liang, X., & Jiang, L. (2017) · Design, Automation & Test in Europe Conference & Exhibition (DATE), 2017 · doi:10.23919/date.2017.7926952 · Cited by 245 · cited 1×
“For example, a defect map or variation distribution [153], [155], [157] of devices in the crossbar can aid the training process by mapping the sensitive cells to defect-free or low variation cells.”
BackgroundEvidences of oxygen-mediated resistive-switching mechanism in TiN\HfO2\Pt cells
Goux, L., Czarnecki, P., Chen, Y. Y., Pantisano, L., Wang, X. P., Degraeve, R., Govoreanu, B., Jurczak, M., Wouters, D. J., & Altimime, L. (2010) · Applied Physics Letters · doi:10.1063/1.3527086 · Cited by 232 · cited 1×
“[49] L.”
Mitigating effects of non-ideal synaptic device characteristics for on-chip learning
Chen, P.-Y., Lin, B., Wang, I.-T., Hou, T.-H., Ye, J., Vrudhula, S., Seo, J.-S., Cao, Y., & Yu, S. (2015) · 2015 IEEE/ACM International Conference on Computer-Aided Design (ICCAD) · doi:10.1109/iccad.2015.7372570 · Cited by 232 · cited 1×
“This behavior has been quantitatively modeled by researchers for RRAMs [115], [116] through the following equations : where P is the number of applied pulses, G + ( G − ) is the conductance trajectory in the increasing (decreasing) direction, G min ( G max ) is the minimum (maximum) conductance,…”
MethodXue, C.-X., Huang, T.-Y., Liu, J.-S., Chang, T.-W., Kao, H.-Y., Wang, J.-H., Liu, T.-W., Wei, S.-Y., Huang, S.-P., Wei, W.-C., et al. (2020) · 2020 IEEE International Solid- State Circuits Conference - (ISSCC) · doi:10.1109/isscc19947.2020.9063078 · Cited by 230
Fault and Error Tolerance in Neural Networks: A Review
Torres-Huitzil, C., & Girau, B. (2017) · IEEE Access · doi:10.1109/access.2017.2742698 · Cited by 228 · cited 1×
“In addition to the aforementioned nonidealities, NVM technologies suffer from other kinds of reliability issues arising from manufacturing defects and faults raising testing challenges [124], [125].”
BackgroundAres: a framework for quantifying the resilience of deep neural networks
Reagen, B., Gupta, U., Pentecost, L., Whatmough, P., Lee, S. K., Mulholland, N., Brooks, D., & Wei, G.-Y. (2018) · Proceedings of the 55th Annual Design Automation Conference · doi:10.1145/3195970.3195997 · Cited by 228 · cited 1×
“To that effect, frameworks, such as Ares [146] and Distiller [147], have been developed in congruity with TensorFlow and PyTorch to enable hardware–software codesign.”
MethodRoadmap for Emerging Materials for Spintronic Device Applications
Hirohata, A., Sukegawa, H., Yanagihara, H., Zutic, I., Seki, T., Mizukami, S., & Swaminathan, R. (2015) · IEEE Transactions on Magnetics · doi:10.1109/tmag.2015.2457393 · Cited by 224 · cited 1×
“Although TMR of 600% has been achieved experimentally for MTJs [46] and theoretical predicting that 1000% can be achievable [62], [63], large-scale demonstrations are still lacking.”
BackgroundA Multi-Functional In-Memory Inference Processor Using a Standard 6T SRAM Array
Kang, M., Gonugondla, S. K., Patil, A., & Shanbhag, N. R. (2018) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2017.2782087 · Cited by 224 · cited 1×
“[16] M.”
TaO x -based resistive switching memories: prospective and challenges
Prakash, A., Jana, D., & Maikap, S. (2013) · Nanoscale Research Letters · doi:10.1186/1556-276x-8-418 · Cited by 223 · cited 1×
“There could be different conductance drifting modes, where the conductance of the device can move toward a maximum, minimum, and intermediate states [121], [122].”
BackgroundKawahara, T., Takemura, R., Miura, K., Hayakawa, J., Ikeda, S., Lee, Y., Sasaki, R., Goto, Y., Ito, K., Meguro, T., et al. (2007) · 2007 IEEE International Solid-State Circuits Conference. Digest of Technical Papers · doi:10.1109/isscc.2007.373503 · Cited by 218 · cited 2×
“[61] T.”
BackgroundYu, S., Gao, B., Fang, Z., Yu, H., Kang, J., & Wong, H.-S. P. (2012) · 2012 International Electron Devices Meeting · doi:10.1109/iedm.2012.6479018 · Cited by 213 · cited 1×
“[119] S.”
Resistive memory device requirements for a neural algorithm accelerator
Agarwal, S., Plimpton, S. J., Hughart, D. R., Hsia, A. H., Richter, I., Cox, J. A., James, C. D., & Marinella, M. J. (2016) · 2016 International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2016.7727298 · Cited by 213 · cited 2×
“Through this approach, weight gradient computations and updates can be performed in situ in resistive crossbars [140].”
BackgroundMethodSpin-Transfer Torque Devices for Logic and Memory: Prospects and Perspectives
Fong, X., Kim, Y., Yogendra, K., Fan, D., Sengupta, A., Raghunathan, A., & Roy, K. (2016) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2015.2481793 · Cited by 211 · cited 1×
“NVM devices are mostly two-terminal devices based on phase-change materials (e.g., PCRAM) [37], oxide-based materials (e.g., RRAMs) [38], or magnetic materials [e.g., spin-transfer torque (STT)-MRAM] [39].”
BackgroundSelf-citationA Retrospective and Prospective View of Approximate Computing [Point of View}
Liu, W., Lombardi, F., & Shulte, M. (2020) · Proceedings of the IEEE · doi:10.1109/jproc.2020.2975695 · Cited by 210 · cited 1×
“The analog nature of computing in NVM crossbars introduces various layers of approximation, which can lead to unique accuracy-energy tradeoffs [32].”
BackgroundRoadmap for Emerging Materials for Spintronic Device Applications
Hirohata, A., Sukegawa, H., Yanagihara, H., Žutić, I., Seki, T., Mizukami, S., & Swaminathan, R. (2015) · White Rose Research Online (University of Leeds, The University of Sheffield, University of York) · Cited by 206
Vortex: variation-aware training for memristor X-bar
Liu, B., Li, H., Chen, Y., Li, X., Wu, Q., & Huang, T. (2015) · Proceedings of the 52nd Annual Design Automation Conference · doi:10.1145/2744769.2744930 · Cited by 197 · cited 1×
“For example, a defect map or variation distribution [153], [155], [157] of devices in the crossbar can aid the training process by mapping the sensitive cells to defect-free or low variation cells.”
BackgroundKim, S., Ishii, M., Lewis, S., Perri, T., BrightSky, M., Kim, W., Jordan, R., Burr, G. W., Sosa, N., Ray, A., et al. (2015) · 2015 IEEE International Electron Devices Meeting (IEDM) · doi:10.1109/iedm.2015.7409716 · Cited by 194 · cited 1×
“There have been several explorations of in-memory computing to perform MVM operations on phase-change memories, both in electronic [41] and photonic domains [42].”
BackgroundMochida, R., Kouno, K., Hayata, Y., Nakayama, M., Ono, T., Suwa, H., Yasuhara, R., Katayama, K., Mikawa, T., & Gohou, Y. (2018) · 2018 IEEE Symposium on VLSI Technology · doi:10.1109/vlsit.2018.8510676 · Cited by 194 · cited 1×
“[96] R.”
8T SRAM Cell as a Multibit Dot-Product Engine for Beyond Von Neumann Computing
Jaiswal, A., Chakraborty, I., Agrawal, A., & Roy, K. (2019) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2019.2929245 · Cited by 191 · cited 1×
“10(a) shows an example of such a macro based on an eight-transistor (8T) SRAM array [14].”
BackgroundSelf-citationAres: A framework for quantifying the resilience of deep neural networks
Reagen, B., Gupta, U., Pentecost, L., Whatmough, P., Lee, S. K., Mulholland, N., Brooks, D., & Wei, G.-Y. (2018) · 2018 55th ACM/ESDA/IEEE Design Automation Conference (DAC) · doi:10.1109/dac.2018.8465834 · Cited by 189
Merced-Grafals, E. J., Dávila, N., Ge, N., Williams, R. S., & Strachan, J. P. (2016) · Nanotechnology · doi:10.1088/0957-4484/27/36/365202 · Cited by 184 · cited 2×
“Second, the high write voltage [103] and multiple programming cycles (programverify approach [45]) result in much higher write cost (energy and latency) than SRAM.”
BackgroundMarinella, M. J., Agarwal, S., Hsia, A., Richter, I., Jacobs-Gedrim, R., Niroula, J., Plimpton, S. J., Ipek, E., & James, C. D. (2018) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2018.2796379 · Cited by 183 · cited 5×
“Second, the high write voltage [103] and multiple programming cycles (programverify approach [45]) result in much higher write cost (energy and latency) than SRAM.”
BackgroundInfluentialBinary convolutional neural network on RRAM
Tang, T., Xia, L., Li, B., Wang, Y., & Yang, H. (2017) · 2017 22nd Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1109/aspdac.2017.7858419 · Cited by 176 · cited 1×
“Due to the intrinsic physics of these NVM devices, they can store multiple states in a highly dense memory with a bit-cell area of ∼ 4 F 2 [27].”
BackgroundSun, X., & Yu, S. (2019) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2019.2933148 · Cited by 170 · cited 6×
“The impact of this undesirable characteristic is studied [116] on CIFAR-10 data set where the training accuracy is severely affected, especially under the effects of write nonlinearity, and a good endurance cycle ( ∼ 10 4 cycles) is necessary for proper training.”
BackgroundMethodInfluentialPhotonic In-Memory Computing Primitive for Spiking Neural Networks Using Phase-Change Materials
Chakraborty, I., Saha, G., & Roy, K. (2019) · Physical Review Applied · doi:10.1103/physrevapplied.11.014063 · Cited by 169
Self-citationPathways to efficient neuromorphic computing with non-volatile memory technologies
Chakraborty, I., Jaiswal, A., Saha, A. K., Gupta, S. K., & Roy, K. (2020) · Applied Physics Reviews · doi:10.1063/1.5113536 · Cited by 168 · cited 2×
“[20] I.”
BackgroundSelf-citationMultilevel resistive switching in Ti/CuxO/Pt memory devices
Wang, S.-Y., Huang, C.-W., Lee, D.-Y., Tseng, T.-Y., & Chang, T.-C. (2010) · Journal of Applied Physics · doi:10.1063/1.3518514 · Cited by 164 · cited 2×
“[53] S.”
Background3D-stackable crossbar resistive memory based on Field Assisted Superlinear Threshold (FAST) selector
Jo, S. H., Kumar, T., Narayanan, S., Lu, W. D., & Nazarian, H. (2014) · 2014 IEEE International Electron Devices Meeting · doi:10.1109/iedm.2014.7046999 · Cited by 162 · cited 3×
“Moreover, field-assisted superlinear threshold (FAST) selector has been reported in [84] with steep turn-on slope (<5 mV/decade) and high ON-current JON (>5 MA/cm).”
BackgroundTraining Deep Convolutional Neural Networks with Resistive Cross-Point Devices
Gokmen, T., Onen, M., & Haensch, W. (2017) · Frontiers in Neuroscience · doi:10.3389/fnins.2017.00538 · Cited by 162 · cited 1×
“To address this problem, researchers have proposed noise management techniques, such as normalization of inputs and rescaling of outputs [159].”
Sengupta, A., & Roy, K. (2017) · Applied Physics Reviews · doi:10.1063/1.5012763 · Cited by 158
Self-citationIelmini, D., Lavizzari, S., Sharma, D., & Lacaita, A. L. (2007) · 2007 IEEE International Electron Devices Meeting · doi:10.1109/iedm.2007.4419107 · Cited by 155 · cited 1×
“In the context of NVM for ML workloads, the drifting effects have been analyzed [116] by a widely used PCM drift model [123] where G 0 is the initial conductance, v is the drift coeffi-cient, and t 0 is the time constant.”
MethodChen, Z., & Liu, B. (2018) · Synthesis Lectures on Artificial Intelligence and Machine Learning · doi:10.2200/s00737ed1v01y201610aim033 · Cited by 154
Variability of resistive switching memories and its impact on crossbar array performance
Chen, A., & Lin, M.-R. (2011) · 2011 International Reliability Physics Symposium · doi:10.1109/irps.2011.5784590 · Cited by 152 · cited 1×
“RRAM technology, however, still needs significant development as it is prone to intercrossbar and intracrossbar variabilities, especially due to the filament formation process [58].”
BackgroundValavi, H., Ramadge, P. J., Nestler, E., & Verma, N. (2018) · 2018 IEEE Symposium on VLSI Circuits · doi:10.1109/vlsic.2018.8502421 · Cited by 150 · cited 1×
“As a result, researchers have explored CMOS macros primarily in the context of binary neural networks [107], [108] where the matrix–vector multiplications are represented using XNOR operations [57], [109] between binary inputs and weights.”
BackgroundFong, X., Gupta, S. K., Mojumder, N. N., Choday, S. H., Augustine, C., & Roy, K. (2011) · International Conference on Simulation of Semiconductor Processes and Devices · doi:10.1109/sispad.2011.6035047 · Cited by 145 · cited 1×
“In addition, the nonlinear I versus V characteristics in spin devices can be modeled using nonequilibrium Green’s function (NEGF) [114].”
MethodSelf-citationAli, T., Polakowski, P., Kühnel, K., Czernohorsky, M., Kämpfe, T., Rudolph, M., Pätzold, B., Lehninger, D., Müller, F., Olivo, R. R., et al. (2019) · Technical digest - International Electron Devices Meeting/Technical digest · doi:10.1109/iedm19573.2019.8993642 · Cited by 145 · cited 1×
“Even though FE materials have been explored as a nonvolatile capacitor for a long time [71], recent devices propose integration of FE layer in the gate-stack of a standard transistor [47], [72], [73].”
BackgroundRENO: a high-efficient reconfigurable neuromorphic computing accelerator design
Liu, X., Mao, M., Liu, B., Li, H., Chen, Y., Li, B., Wang, Y., Jiang, H., Barnell, M., Wu, Q., et al. (2015) · Proceedings of the 52nd Annual Design Automation Conference · doi:10.1145/2744769.2744900 · Cited by 142 · cited 2×
“Resistive crossbars have been used to build special-purpose ML accelerators, such as SPINDLE [28], RENO [135], PRIME [136], ISAAC [29] for convolutional neural networks (CNNs), and MLPs.”
MethodRxNN: A Framework for Evaluating Deep Neural Networks on Resistive Crossbars
Jain, S., Sengupta, A., Roy, K., & Raghunathan, A. (2021) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2020.3000185 · Cited by 137 · cited 2×
“Fast crossbar model (FCM) [130] is a weight-transformation-based approach to enable fast analytical modeling of resistive crossbars.”
BackgroundMethodSelf-citationStatistical Fluctuations in HfOx Resistive-Switching Memory: Part II—Random Telegraph Noise
Ambrogio, S., Balatti, S., Cubeta, A., Calderoni, A., Ramaswamy, N., & Ielmini, D. (2014) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2014.2330202 · Cited by 134 · cited 2×
“The origin of these variations has been studied in detailed for various technologies [163], [164].”
BackgroundSneak-path constraints in memristor crossbar arrays
Cassuto, Y., Kvatinsky, S., & Yaakobi, E. (2013) · IEEE International Symposium on Information Theory (ISIT) · doi:10.1109/isit.2013.6620207 · Cited by 128 · cited 1×
“However, due to merged read and write paths, writing into these devices may cause sneak path issues [64].”
BackgroundGopalakrishnan, K., Shenoy, R. S., Rettner, C. T., Virwani, K., Bethune, D. S., Shelby, R. M., Burr, G. W., Kellock, A., King, R. S., Nguyen, K., et al. (2010) · 2010 Symposium on VLSI Technology · doi:10.1109/vlsit.2010.5556229 · Cited by 126 · cited 2×
“V characteristics that can be adopted in 1S-1R crossbars [80].”
BackgroundGENIEx: A Generalized Approach to Emulating Non-Ideality in Memristive Xbars using Neural Networks
Chakraborty, I., Ali, M., Kim, D. E., Ankit, A., & Roy, K. (2020) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18072.2020.9218688 · Cited by 124 · cited 1×
“D data fitting of nonideal output current I nonideal as Such a regression model can be extended to a multilayer perceptron (MLP) network to generalize the behavior of the resistive crossbar [132].”
BackgroundSelf-citationReduction and IR-drop compensations techniques for reliable neuromorphic computing systems
Liu, B., Li, H., Chen, Y., Li, X., Huang, T., Wu, Q., & Barnell, M. (2014) · 2014 IEEE/ACM International Conference on Computer-Aided Design (ICCAD) · doi:10.1109/iccad.2014.7001330 · Cited by 124 · cited 1×
“One such technique [154] performs linear algebraic methods to reduce the crossbar size by removing redundant elements to minimize the impact of wire resistances.”
MethodParasitic Effect Analysis in Memristor-Array-Based Neuromorphic Systems
Jeong, Y., Zidan, M. A., & Lu, W. D. (2018) · IEEE Transactions on Nanotechnology · doi:10.1109/tnano.2017.2784364 · Cited by 115 · cited 2×
“The impact of parasitics can be also modeled using modified nodal analysis, as adopted in [131].”
MethodXue, C.-X., Chen, W.-H., Liu, J.-S., Li, J.-F., Lin, W.-Y., Lin, W.-E., Wang, J.-H., Wei, W.-C., Huang, T.-Y., Chang, T.-W., et al. (2020) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2019.2951363 · Cited by 113 · cited 2×
“Various ADC architectures and precisions have been adopted in memristive crossbar MVM units [29], [30], [88].”
BackgroundMethodAccumulative Polarization Reversal in Nanoscale Ferroelectric Transistors
Mulaosmanovic, H., Mikolajick, T., & Slesazeck, S. (2018) · ACS Applied Materials & Interfaces · doi:10.1021/acsami.8b08967 · Cited by 107 · cited 1×
“As a result, FeFETs have shown immense promise as neuromimetic devices [74]–[77].”
BackgroundImprovement of data retention in HfO2/Hf 1T1R RRAM cell under low operating current
Chen, Y. Y., Komura, M., Degraeve, R., Govoreanu, B., Goux, L., Fantini, A., Raghavan, N., Clima, S., Zhang, L., Belmonte, A., et al. (2013) · 2013 IEEE International Electron Devices Meeting · doi:10.1109/iedm.2013.6724598 · Cited by 102 · cited 1×
“There could be different conductance drifting modes, where the conductance of the device can move toward a maximum, minimum, and intermediate states [121], [122].”
BackgroundConductive-bridging random access memory: challenges and opportunity for 3D architecture
Jana, D., Roy, S., Panja, R., Dutta, M., Rahaman, S. Z., Mahapatra, R., & Maikap, S. (2015) · Nanoscale Research Letters · doi:10.1186/s11671-015-0880-9 · Cited by 102 · cited 1×
“Another class of devices formed using the same structure by simply replacing the oxides by conductive elements constitutes conductive bridge RAMs or CBRAMs [52].”
BackgroundSelf-citationProgramming Protocol Optimization for Analog Weight Tuning in Resistive Memories
Gao, L., Chen, P.-Y., & Yu, S. (2015) · IEEE Electron Device Letters · doi:10.1109/led.2015.2481819 · Cited by 100 · cited 1×
“For offline trained neural networks, the conductances of the NVM devices in the crossbar are predetermined and, hence, can be reliably written using the popular read– verify–write mechanism [117].”
MethodRacetrack memory cell array with integrated magnetic tunnel junction readout
Annunziata, A. J., Gaidis, M. C., Thomas, L., Chien, C.-H., Hung, C. C., Chevalier, P., O'Sullivan, E. J., Hummel, J. P., Joseph, E. A., Zhu, Y., et al. (2011) · Technical digest - International Electron Devices Meeting/Technical digest · doi:10.1109/iedm.2011.6131604 · Cited by 96 · cited 1×
“Moreover, SHE-MTJs based on DWMs have issues with domain-wall stability, which causes variability in the intermediate states, making it unsuitable for large-scale crossbar implementations [70].”
BackgroundAgarwal, S., Quach, T.-T., Parekh, O., Hsia, A. H., DeBenedictis, E. P., James, C. D., Marinella, M. J., & Aimone, J. B. (2016) · Frontiers in Neuroscience · doi:10.3389/fnins.2015.00484 · Cited by 96 · cited 2×
“the NVM devices at the crosspoint [101]. crosspoint) to a detailed description of peripheral circuits required for proper operation of the crossbar as compute cores in DNN accelerators.”
BackgroundAnkit, A., Sengupta, A., Panda, P., & Roy, K. (2017) · Proceedings of the 54th Annual Design Automation Conference 2017 · doi:10.1145/3061639.3062311 · Cited by 92 · cited 2×
“[137] A.”
MethodSelf-citationDeterministic Domain Wall Motion Orthogonal To Current Flow Due To Spin Orbit Torque
Bhowmik, D., Nowakowski, M., You, L., Lee, O., Keating, D., Wong, M., Jeffrey Bokor, A., & Salahuddin, S. (2015) · Scientific Reports · doi:10.1038/srep11823 · Cited by 91 · cited 2×
“While the stability of the Neel wall is aided by the Dzyaloshinskii–Moriya interaction [67], [68], the Bloch walls need external magnetic fields for proper operation [69].”
MethodThe Current-Steering DAC [A Circuit for All Seasons]
Razavi, B. (2018) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2017.2771102 · Cited by 90 · cited 1×
“Different variants of DAC utilize charge, current, and voltage multiplication or division using switched-capacitor circuits, current steering circuits, and resistor ladders, respectively, to generate the analog output [87], [90].”
BackgroundTechnology Aware Training in Memristive Neuromorphic Systems for Nonideal Synaptic Crossbars
Chakraborty, I., Roy, D., & Roy, K. (2018) · IEEE Transactions on Emerging Topics in Computational Intelligence · doi:10.1109/tetci.2018.2829919 · Cited by 89 · cited 3×
“Such retraining techniques can be further improved by modifying the gradient descent algorithm [34] to account for nonidealities. b) Compensation through mapping: An alternative approach is to use mapping algorithms [89] for weight to conductance conversion by accounting for computational errors…”
BackgroundMethodSelf-citationPANTHER: A Programmable Architecture for Neural Network Training Harnessing Energy-Efficient ReRAM
Ankit, A., Hajj, I. E., Chalamalasetti, S. R., Agarwal, S., Marinella, M., Foltin, M., Strachan, J. P., Milojicic, D., Hwu, W.-M., & Roy, K. (2020) · IEEE Transactions on Computers · doi:10.1109/tc.2020.2998456 · Cited by 88 · cited 4×
“To represent both positive and negative weights in the same device, the range of weight, RON to ROFF, is symmetrically centered around (RON + ROFF)/2 [102].”
BackgroundMethodInfluentialSelf-citationSPINDLE: SPINtronic deep learning engine for large-scale neuromorphic computing
Ramasubramanian, S. G., Venkatesan, R., Sharad, M., Roy, K., & Raghunathan, A. (2014) · Proceedings of the 2014 international symposium on Low power electronics and design · doi:10.1145/2627369.2627625 · Cited by 85 · cited 2×
“This promise has fueled great interest over the past decade in designing systems that can leverage the high computing efficiency as well as the high-density storage of crossbar-based MVM units [28]–[30].”
BackgroundMethodSelf-citationA Vision for All-Spin Neural Networks: A Device to System Perspective
Sengupta, A., & Roy, K. (2016) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2016.2615312 · Cited by 75 · cited 2×
“…memory (NVM) technologies [19], [20], such as phase-change memory (PCM) [21], resistive random access memory (RRAM) [22], [23], and spintronics [24], offer immense promise as an alternative to CMOS due to their high storage density and the ability to perform massively parallel in situ MVM…”
BackgroundSelf-citationKim, Y., Zhang, Y., & Li, P. (2015) · ACM Journal on Emerging Technologies in Computing Systems · doi:10.1145/2700234 · Cited by 74 · cited 2×
“Past research has also proposed using memristive crossbars for spiking neural networks [137], [138] and the Boltzmann machines [139].”
MethodZhao, M., Wu, H., Gao, B., Sun, X., Liu, Y., Yao, P., Xi, Y., Li, X., Zhang, Q., Wang, K., et al. (2018) · 2018 IEEE International Electron Devices Meeting (IEDM) · doi:10.1109/iedm.2018.8614664 · Cited by 71 · cited 1×
“As resistive crossbars are primarily associated with analog weight updates, the strength of updates can degrade with the number of programming pulses applied [120], as shown in Fig.”
BackgroundGimenez, S. P. (2016) · Synthesis lectures on emerging engineering technologies · doi:10.2200/s00704ed1v01y201602eet007 · Cited by 67
Analog/Mixed-Signal Hardware Error Modeling for Deep Learning Inference
Rekhi, A. S., Zimmer, B., Nedovic, N., Liu, N., Venkatesan, R., Wang, M., Khailany, B., Dally, W. J., & Gray, C. T. (2019) · Proceedings of the 56th Annual Design Automation Conference 2019 · doi:10.1145/3316781.3317770 · Cited by 67 · cited 3×
“workloads to be mapped to such hardware fabrics and evaluated considering the impact of nonidealities. precision [36], we can leverage the error-resiliency of neural networks to strike a balance between the accuracy and energy of NVM crossbar systems by varying the ADC precision.”
BackgroundMethodInfluentialMurmann, B. (2016) · IEEE Communications Magazine · doi:10.1109/mcom.2016.7452270 · Cited by 66 · cited 2×
“Due to a large number of comparators, the energy consumption of a flash ADC is much larger than the same-precision SAR ADC [95].”
BackgroundA Deep Neural Network Accelerator Based on Tiled RRAM Architecture
Wang, Q., Wang, X., Lee, S. H., Meng, F.-H., & Lu, W. D. (2019) · 2019 IEEE International Electron Devices Meeting (IEDM) · doi:10.1109/iedm19573.2019.8993641 · Cited by 66 · cited 1×
“[99] Q.”
A Phase Change Memory Compact Model for Multilevel Applications
Ventrice, D., Fantini, P., Redaelli, A., Pirovano, A., Benvenuti, A., & Pellizzer, F. (2007) · IEEE Electron Device Letters · doi:10.1109/led.2007.907288 · Cited by 64 · cited 1×
“V behavior has been previously studied [112], [113].”
MethodParallel Architecture With Resistive Crosspoint Array for Dictionary Learning Acceleration
Kadetotad, D., Xu, Z., Mohanty, A., Chen, P.-Y., Lin, B., Ye, J., Vrudhula, S., Yu, S., Cao, Y., & Seo, J.-S. (2015) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2015.2426495 · Cited by 63 · cited 1×
“Such parallel writing schemes can be adopted entirely in the time domain as well [104].”
BackgroundToward on-chip acceleration of the backpropagation algorithm using nonvolatile memory
Narayanan, P., Fumarola, A., Sanches, L. L., Hosokawa, K., Lewis, S. C., Shelby, R. M., & Burr, G. W. (2017) · IBM Journal of Research and Development · doi:10.1147/jrd.2017.2716579 · Cited by 63 · cited 1×
“The weight change in a cell depends on the duration and the amplitude of the pulses applied on the corresponding row and column, respectively, thereby realizing an analog weight update operation [103], [141].”
BackgroundCxDNN: Hardware-software Compensation Methods for Deep Neural Networks on Resistive Crossbar Systems
Jain, S., & Raghunathan, A. (2019) · ACM Transactions on Embedded Computing Systems · doi:10.1145/3362035 · Cited by 63
Self-citationResistive Random Access Memory (RRAM)
Yu, S. (2016) · Synthesis lectures on emerging engineering technologies · doi:10.2200/s00681ed1v01y201510eet006 · Cited by 57
Improvement of data retention in HfO2 / Hf 1T1R RRAM cell under low operating current
Chen, Y. Y., Komura, M., Degraeve, R., Govoreanu, B., Goux, L., Fantini, A., Raghavan, N., Clima, S., Zhang, L., Belmonte, A., et al. (2013) · Cited by 55
Newton: Gravitating Towards the Physical Limits of Crossbar Acceleration
Nag, A., Balasubramonian, R., Srikumar, V., Walker, R., Shafiee, A., Strachan, J. P., & Muralimanohar, N. (2018) · IEEE Micro · doi:10.1109/mm.2018.053631140 · Cited by 55
Cross-point memory design challenges and survey of selector device characteristics
Peng, X., Madler, R., Chen, P.-Y., & Yu, S. (2017) · Journal of Computational Electronics · doi:10.1007/s10825-017-1062-z · Cited by 53 · cited 1×
“At high voltages, the selectors are turned on, and current passes through the memristive device [79], [86].”
BackgroundNi, L., Wang, Y., Yu, H., Yang, W., Weng, C., & Zhao, J. (2016) · 2016 21st Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1109/aspdac.2016.7428024 · Cited by 52 · cited 3×
“These peripheral circuits within MVM units, particularly the ADCs, have been shown to account for a large fraction (up to ∼80%) of the total energy consumption of the MVM unit [35].”
BackgroundInfluentialCrossbar-Aware Neural Network Pruning
Liang, L., Deng, L., Zeng, Y., Hu, X., Ji, Y., Ma, X., Li, G., & Xie, Y. (2018) · IEEE Access · doi:10.1109/access.2018.2874823 · Cited by 52 · cited 1×
“Previous works have focused on structured sparsity [165]–[167].”
BackgroundThe Transimpedance Amplifier [A Circuit for All Seasons]
Razavi, B. (2019) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2018.2881860 · Cited by 52 · cited 3×
“A common circuit technique to convert analog current to voltage is the TIA [92].”
MethodInfluentialTxSim: Modeling Training of Deep Neural Networks on Resistive Crossbar Systems
Roy, S., Sridharan, S., Jain, S., & Raghunathan, A. (2021) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2021.3063543 · Cited by 52 · cited 1×
“size and RON/ROFF ratio (reprinted from [151]).”
Self-citationWoo, J., & Yu, S. (2019) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2019.2917764 · Cited by 48 · cited 4×
“On the other hand, the applied voltage to 1TS-1R cell beyond TS threshold voltage is fully transferred to the RRAM due to the fast slope that is at most 5 mV/decade [79].”
BackgroundInfluentialHardware-Software Co-Design for an Analog-Digital Accelerator for Machine Learning
Ambrosi, J., Ankit, A., Antunes, R., Chalamalasetti, S. R., Chatterjee, S., Hajj, I. E., Fachini, G. J. D. A., Faraboschi, P., Foltín, M., Huang, S., et al. (2018) · IEEE International Conference on Rebooting Computing (ICRC) · doi:10.1109/icrc.2018.8638612 · Cited by 47 · cited 1×
“2) Performance Modeling: Energy and Latency: The energy and latency metrics of NVM crossbar-based architectures can be estimated through performance modeling tools [30], [31].”
MethodSelf-citationDemonstration of Reliable Triple-Level-Cell (TLC) Phase-Change Memory
Stanisavljević, M., Pozidis, H., Athmanathan, A., Papandreou, N., Mittelholzer, T., & Eleftheriou, E. S. (2016) · IEEE International Memory Workshop (IMW) · doi:10.1109/imw.2016.7495263 · Cited by 45 · cited 1×
“[44] M.”
Transition between Bloch and Néel Walls
Torok, E. J., Olson, A. L., & Oredson, H. N. (1965) · Journal of Applied Physics · doi:10.1063/1.1714317 · Cited by 41 · cited 1×
“Devices with perpendicular magnetic anisotropy (PMA) exhibit Neel wall in narrow nanostrips and Bloch wall in wider nanostrips [66].”
BackgroundFerroelectric FET Based In-Memory Computing for Few-Shot Learning
Laguna, A. F., Yin, X., Reis, D. A., Niemier, M. T., & Hu, X. S. (2019) · ACM Great Lakes Symposium on VLSI (GLSVLSI) · doi:10.1145/3299874.3319450 · Cited by 40 · cited 1×
“These properties can be easily ported for ML workloads [78], as discussed in this article.”
BackgroundZmora, N., Jacob, G., & Novik, G. (2018) · Zenodo (CERN European Organization for Nuclear Research) · doi:10.5281/zenodo.1297430 · Cited by 40
TraNNsformer: Neural network transformation for memristive crossbar based neuromorphic system design
Ankit, A., Sengupta, A., & Roy, K. (2017) · 2017 IEEE/ACM International Conference on Computer-Aided Design (ICCAD) · doi:10.1109/iccad.2017.8203823 · Cited by 39 · cited 2×
“Previous works have focused on structured sparsity [165]–[167].”
BackgroundSelf-citationNervanaSystems/distiller v0.3.2
Zmora, N., Jacob, G., Bar, Zlotnik, L., Novik, G., Barad, H., Chen, Y.-S., Muchsel, R., 103yiran, Tacker-Oh, et al. (2018) · Zenodo (CERN European Organization for Nuclear Research) · doi:10.5281/zenodo.1297429 · Cited by 39
Through the Looking Glass?The 2015 Edition: Trends in Solid-State Circuits from ISSCC
Narendra, S., Fujino, L. C., & Smith, K. C. (2015) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2014.2375071 · Cited by 38 · cited 3×
“[134] S.”
BackgroundEmerging Memory Technologies: Design, Architecture, and Applications
Xie, Y. (2013) · CERN Document Server (European Organization for Nuclear Research) · doi:10.1007/978-1-4419-9551-3 · Cited by 38
Kim, W. G., Lee, H. M., Kim, B. Y., Jung, K. H., Seong, T. G., Kim, S., Jung, H. C., Kim, H. J., Yoo, J. H., Lee, H. D., et al. (2014) · 2014 Symposium on VLSI Technology (VLSI-Technology): Digest of Technical Papers · doi:10.1109/vlsit.2014.6894405 · Cited by 37 · cited 2×
“Such a threshold switching phenomenon can be implemented in the metal–insulator transition (MIT) materials, such as NbO 2 [82].”
BackgroundDRC2: Dynamically Reconfigurable Computing Circuit based on memory architecture
Akyel, K. C., Charles, H., Mottin, J., Giraud, B., Suraci, G., Thuries, S., & Noël, J.-P. (2016) · IEEE International Conference on Rebooting Computing (ICRC) · doi:10.1109/icrc.2016.7738698 · Cited by 37 · cited 2×
“2) CMOS MVM Macros: It is worth mentioning that MVM units can also be constructed using CMOS-based SRAM cells [13]–[15], [106].”
BackgroundGroup Scissor: Scaling Neuromorphic Computing Design to Large Neural Networks
Wang, Y., Wen, W., Liu, B., Chiarulli, D. M., & Li, H. H. (2017) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1145/3061639.3062256 · Cited by 37 · cited 1×
“[166] Y.”
Self‐formed Schottky barrier induced selector‐less RRAM for cross‐point memory applications
Park, S., Jung, S., Siddik, M., Jo, M., Park, J., Kim, S., Lee, W., Shin, J., Lee, D., Choi, G., et al. (2012) · physica status solidi (RRL) - Rapid Research Letters · doi:10.1002/pssr.201206382 · Cited by 36 · cited 2×
“[54] S.”
BackgroundA cross point Cu-ReRAM with a novel OTS selector for storage class memory applications
Yasuda, S., Ohba, K., Mizuguchi, T., Sei, H., Shimuta, M., Aratani, K., 椎本, 恒., Yamamoto, T., Sone, T., 野々口, 誠., et al. (2017) · Symposium on VLSI Technology · doi:10.23919/vlsit.2017.7998189 · Cited by 34 · cited 3×
“D integration than a transistor; thus, one-selector–one-memristive (1S-1R) device arrays were reported with different types of selector devices, such as in [80]–[82], [84], and [85].”
BackgroundKim, Y., Kim, H., Ahn, D., & Kim, J. (2018) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1145/3218603.3218605 · Cited by 33 · cited 2×
“PDAC = β(2 /N + 1)V fclk [142].”
A Tale of Two ADCs: Pipelined Versus SAR
Razavi, B. (2015) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2015.2442372 · Cited by 32 · cited 1×
“To that effect, various ADC architectures have been adopted in memristive crossbars, including successive approximation register (SAR)-based ADC and flash ADC [93], [94].”
MethodBiologically Plausible Ferroelectric Quasi-Leaky Integrate and Fire Neuron
Dutta, S., Saha, A. K., Panda, P., Chakraborty, W., Gomez, J. T., Khanna, A., Gupta, S. K., Roy, K., & Datta, S. (2019) · Symposium on VLSI Technology · doi:10.23919/vlsit.2019.8776487 · Cited by 32 · cited 1×
“[75] S.”
Self-citationModeling and design analysis of 3D vertical resistive memory — A low cost cross-point architecture
Xu, C., Niu, D., Yu, S., & Xie, Y. (2014) · Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1109/aspdac.2014.6742992 · Cited by 29 · cited 1×
“[118] C.”
The Flash ADC [A Circuit for All Seasons]
Razavi, B. (2017) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2017.2712998 · Cited by 29 · cited 3×
“To that effect, various ADC architectures have been adopted in memristive crossbars, including successive approximation register (SAR)-based ADC and flash ADC [93], [94].”
BackgroundMethodInfluentialAgrawal, A., Lee, C., & Roy, K. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1907.00285 · Cited by 28 · cited 2×
“…be of three types: 1) retraining of weights for error compensation [33], [34]; 2) mapping algorithms to reduce the impact of nonidealities [153]–[156]; and 3) hardware compensation [33]. a) Compensation through retraining: When large-scale DNNs are mapped to nonideal crossbars using the…”
BackgroundSelf-citationTraNNsformer: Neural Network Transformation for memristive crossbar based neuromorphic system designPreprint
Ankit, A., Sengupta, A., & Roy, K. (2017) · arXiv (Cornell University) · doi:10.5555/3199700.3199771 · Cited by 27
Self-citationNeural network accelerator design with resistive crossbars: Opportunities and challenges
Jain, S., Ankit, A., Chakraborty, I., Gokmen, T., Rasch, M., Haensch, W., Roy, K., & Raghunathan, A. (2019) · IBM Journal of Research and Development · doi:10.1147/jrd.2019.2947011 · Cited by 27 · cited 2×
“[149] S.”
Self-citationWoo, J., Lee, D., Cha, E., Lee, S., Park, S., & Hwang, H. (2013) · IEEE Electron Device Letters · doi:10.1109/led.2013.2285583 · Cited by 26 · cited 1×
“V slope to turn on the selector associated with an exponential current increase [81].”
BackgroundModeling and Comparative Analysis of Hysteretic Ferroelectric and Anti-ferroelectric FETs
Saha, A. K., & Gupta, S. K. (2018) · Conference digest · doi:10.1109/drc.2018.8442136 · Cited by 21 · cited 1×
“[76] A.”
Cao, N., Chang, M., & Raychowdhury, A. (2019) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2019.2935533 · Cited by 20 · cited 2×
“[91] N.”
MethodThe Impact of Ferroelectric FETs on Digital and Analog Circuits and Architectures
Chen, X., Sun, X., Wang, P., Datta, S., Hu, X. S., Yin, X., Jerry, M., Yu, S., Laguna, A. F., Ni, K., et al. (2019) · IEEE Design and Test · doi:10.1109/mdat.2019.2944094 · Cited by 20 · cited 1×
“Even though FE materials have been explored as a nonvolatile capacitor for a long time [71], recent devices propose integration of FE layer in the gate-stack of a standard transistor [47], [72], [73].”
BackgroundA Cross Point Cu-ReRAM with a Novel OTS Selector for Storage Class Memory Applications
保田, 周., 大場, 和., 水口, 徹., 清, 宏., 紫牟田, 雅., 荒谷, 勝., 椎本, 恒., 山元, 哲., 曽根, 威., 野々口, 誠., et al. (2017) · Cited by 19
The R-2R and C-2C Ladders [A Circuit for All Seasons]
Razavi, B. (2019) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2019.2922886 · Cited by 14 · cited 1×
“Different variants of DAC utilize charge, current, and voltage multiplication or division using switched-capacitor circuits, current steering circuits, and resistor ladders, respectively, to generate the analog output [87], [90].”
BackgroundRebooting Computing: The Challenges for Test and Reliability
Bosio, A., Hamdioui, S., O’Connor, I., Rodrigues, G. S., Lima, F. K., Vatajelu, E. I., Natale, G. D., Anghel, L., Nagarajan, S., & Fieback, M. (2019) · Proceedings · doi:10.1109/dft.2019.8875270 · Cited by 10 · cited 1×
“In addition to the aforementioned nonidealities, NVM technologies suffer from other kinds of reliability issues arising from manufacturing defects and faults raising testing challenges [124], [125].”
BackgroundTraining large-scale ANNs on simulated resistive crossbar arraysPreprint
Rasch, M. J., Gokmen, T., & Hänsch, W. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1906.02698 · Cited by 9
Bojnordi, M. N., & İpek, E. (2017) · doi:10.1109/e3s.2017.8246178 · Cited by 6
Well-Posed Verilog-A Compact Model for Phase Change Memory
Kulkarni, S., Kadetotad, D. V., Seo, J.-S., & Rajendran, B. (2018) · International Conference on Simulation of Semiconductor Processes and Devices · doi:10.1109/sispad.2018.8551667 · Cited by 2 · cited 2×
“[113] S.”
Background
Research works printed but not linked: 3
- agarwal, 2020, CrossSim
- han, 2015, Learning both weights and connections for efficient neural network, Proc Adv Neural Inf Process Syst, 1135
- 2015, 2015 ITRS 2 0 Official Publication zip
Cited by
Hardware implementation of memristor-based artificial neural networks
Aguirre, F., Sebastian, A., Le Gallo, M., Song, W., Wang, T., Yang, J. J., Lu, W., Chang, M.-F., Ielmini, D., Yang, Y., et al. (2024) · Nature Communications · doi:10.1038/s41467-024-45670-9 · Cited by 564
Self-citationIn-memory computing with emerging memory devices: Status and outlook
Mannocci, P., Farronato, M., Lepri, N., Cattaneo, L., Glukhov, A., Sun, Z., & Ielmini, D. (2023) · APL Machine Learning · doi:10.1063/5.0136403 · Cited by 191
Hänsch, W., Raghunathan, A., Roy, K., Chakrabarti, B., Phatak, C., Wang, C., & Guha, S. (2022) · Advanced Materials · doi:10.1002/adma.202204944 · Cited by 167
Self-citationMachine Learning Solutions for the Security of Wireless Sensor Networks: A Review
Ghadi, Y. Y., Mazhar, T., Shloul, T. A., Shahzad, T., Salaria, U. A., Ahmed, A., & Hamam, H. (2024) · IEEE Access · doi:10.1109/access.2024.3355312 · Cited by 155
“Furthermore, it has come to our attention that specific authors have implemented complex ML algorithms to enhance accuracy and efficiency without adequately considering the hardware requirements [123].”
BackgroundBio‐Inspired 3D Artificial Neuromorphic Circuits
Liu, X., Wang, F., Su, J., Zhou, Y., & Ramakrishna, S. (2022) · Advanced Functional Materials · doi:10.1002/adfm.202113050 · Cited by 91
Metadata
- Authors
- Chakraborty, I., Ali, M., Ankit, A., Jain, S., Roy, S., Sridharan, S., Agrawal, A., Raghunathan, A., & Roy, K.
- Affiliations
- Purdue University West Lafayette (US)
- IBM (United States) (US)
- ORCID
- Venue
- Proceedings of the IEEE
- Published
- 2020-12
- Volume
- 108
- Issue
- 12
- Pages
- 2276-2310
- ISSN
- 0018-9219 (ISSN-L)
- 1558-2256
- OpenAlex
- W3042468132
- Cited by
- 140
- References
- 178
- Citations by year
- 2026: 18 · 2025: 31 · 2024: 27 · 2023: 20 · 2022: 22 · 2021: 18 · 2020: 3
- FWCI
- 5.41
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- Proceedings of the IEEE (publishedVersion)
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Memory and Neural Computing
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Low-power high-performance VLSI design
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Keywords
- in-memory computing, machine learning hardware acceleration, matrix-vector multiplication, analog computing, approximate computing, analog-to-digital conversion, spatial architecture, deep neural network inference
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.