XNOR-RRAM: A scalable and parallel resistive synaptic architecture for binary neural networks
The DOI matches its Crossref record; 18 of the 20 references it cites were found in a registry or an index.
Abstract
Recent advances in deep learning have shown that Binary Neural Networks (BNNs) are capable of providing a satisfying accuracy on various image datasets with significant reduction in computation and memory cost. With both weights and activations binarized to +1 or -1 in BNNs, the high-precision multiply-and-accumulate (MAC) operations can be replaced by XNOR and bit-counting operations. In this work, we propose a RRAM synaptic architecture (XNOR-RRAM) with a bit-cell design of complementary word lines that implements equivalent XNOR and bit-counting operation in a parallel fashion. For large-scale matrices in fully connected layers or when the convolution kernels are unrolled in multiple channels, the array partition is necessary. Multi-level sense amplifiers (MLSAs) are employed as the intermediate interface for accumulating partial weighted sum. However, a low bit-level MLSA and intrinsic offset of MLSA may degrade the classification accuracy. We investigate the impact of sensing offsets on classification accuracy and analyze various design options with different sub-array sizes and sensing bit-levels. Experimental results with RRAM models and 65nm CMOS PDK show that the system with 128×128 sub-array size and 3-bit MLSA can achieve accuracies of 98.43% for MLP on MNIST and 86.08% for CNN on CIFAR-10, showing 0.34% and 2.39% degradation respectively compared to the accuracies of ideal BNN algorithms. The projected energy-efficiency of XNOR-RRAM is 141.18 TOPS/W, showing ~33X improvement compared to the conventional RRAM synaptic architecture with sequential row-by-row read-out.
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
“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 [158]
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
XNOR-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
Eyeriss: 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
Courbariaux, M., Hubara, I., Soudry, D., El-Yaniv, R., & Bengio, Y. (2016) · arXiv · doi:10.48550/arxiv.1602.02830 · Cited by 2,106
Quantizing for minimum distortion
Max, J. (1960) · IEEE Transactions on Information Theory · doi:10.1109/tit.1960.1057548 · Cited by 2,038
14.5 Eyeriss: An energy-efficient reconfigurable accelerator for deep convolutional neural networks
Chen, Y.-H., Krishna, T., Emer, J., & Sze, V. (2016) · 2016 IEEE International Solid-State Circuits Conference (ISSCC) · doi:10.1109/isscc.2016.7418007 · Cited by 713
Chen, P.-Y., Peng, X., & Yu, S. (2017) · 2017 IEEE International Electron Devices Meeting (IEDM) · doi:10.1109/iedm.2017.8268337 · Cited by 435
Self-citationMemristor Crossbar-Based Neuromorphic Computing System: A Case Study
Hu, M., Li, H., Chen, Y., Wu, Q., Rose, G. S., & Linderman, R. W. (2014) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2013.2296777 · Cited by 414
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
Self-citationBinary neural network with 16 Mb RRAM macro chip for classification and online training
Yu, S., Li, Z., Chen, P.-Y., Wu, H., Gao, B., Wang, D., Wu, W., & Qian, H. (2016) · 2016 IEEE International Electron Devices Meeting (IEDM) · doi:10.1109/iedm.2016.7838429 · Cited by 219
Self-citation
Cited by
Neuro-Inspired Computing With Emerging Nonvolatile Memorys
Yu, S. (2018) · Proceedings of the IEEE · doi:10.1109/jproc.2018.2790840 · Cited by 1,170
“With this principle, there are a few simulation works to explore the binary neural network with eNVMs [133]–[135].”
BackgroundSelf-citationNeuro-inspired computing chips
Zhang, W., Gao, B., Tang, J., Yao, P., Yu, S., Chang, M.-F., Yoo, H.-J., Qian, H., & Wu, H. (2020) · Nature Electronics · doi:10.1038/s41928-020-0435-7 · Cited by 892
“To overcome the problems of sneak paths and variability, a row-by-row pseudo-crossbar array 64 (Fig.”
BackgroundSelf-citationHardware 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
“In binarized networks, ADC can be replaced by a 1-bit comparator 96 orADC-like multi-level sense ampli fi er 158 .”
Compute-in-Memory Chips for Deep Learning: Recent Trends and Prospects
Yu, S., Jiang, H., Huang, S., Peng, X., & Lu, A. (2021) · IEEE Circuits and Systems Magazine · doi:10.1109/mcas.2021.3092533 · Cited by 391
Self-citationIn situ training of feed-forward and recurrent convolutional memristor networks
Wang, Z., Li, C., Lin, P., Rao, M., Nie, Y., Song, W., Qiu, Q., Li, Y., Yan, P., Strachan, J. P., et al. (2019) · Nature Machine Intelligence · doi:10.1038/s42256-019-0089-1 · Cited by 307
“Theoretical investigations of how the convolutional layers can be mapped to memristor crossbars 41 and memristor CNNs of binary weights 43 have been conducted, with the former experimentally verified on HfO x memristors 25,44 .”
MethodLarge-Scale Neuromorphic Spiking Array Processors: A Quest to Mimic the Brain
Thakur, C. S., Molin, J. L., Cauwenberghs, G., Indiveri, G., Kumar, K., Qiao, N., Schemmel, J., Wang, R., Chicca, E., Olson Hasler, J., et al. (2018) · Frontiers in Neuroscience · doi:10.3389/fnins.2018.00891 · Cited by 288
“For example, image/digit recognition (Seo et al., 2015; Sun et al., 2018) and language modeling have been demonstrated by employing a number of multiple arrays of this chip, where the inter-array communication/computation are performed in digital signals.”
BackgroundSelf-citationA Review of Binarized Neural Networks
Simons, T., & Lee, D.-J. (2019) · Electronics · doi:10.3390/electronics8060661 · Cited by 219
“RRAM is an emerging technology and is an appealing platform for BNN designs due to its compact operation at the bit level [59,60,77,81,82].”
Background90% yield production of polymer nano-memristor for in-memory computing
Zhang, B., Chen, W., Zeng, J., Fan, F., Gu, J., Chen, X., Yan, L., Xie, G., Liu, S., Yan, Q., et al. (2021) · Nature Communications · doi:10.1038/s41467-021-22243-8 · Cited by 218
Research Progress on Memristor: From Synapses to Computing Systems
Yang, X., Taylor, B., Wu, A., Chen, Y., & Chua, L. O. (2022) · IEEE Transactions on Circuits and Systems I: Regular Papers · doi:10.1109/tcsi.2022.3159153 · Cited by 200
Analog architectures for neural network acceleration based on non-volatile memory
Xiao, T. P., Bennett, C. H., Feinberg, B., Agarwal, S., & Marinella, M. J. (2020) · Applied Physics Reviews · doi:10.1063/1.5143815 · Cited by 193
Yin, S., Sun, X., Yu, S., & Seo, J.-S. (2020) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2020.3015178 · Cited by 171
“On the other hand, compared with nonlinear quantization schemes [23], [31], the proposed confined/linear quantization scheme simplifies the ensuing accumulation of ADC outputs (partial sums) and also increases the smallest Vref difference for the adjacent sense amplifiers in the flash ADC.”
BackgroundMethodResultInfluentialSelf-citationSun, X., & Yu, S. (2019) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2019.2933148 · Cited by 170
“The prior work [23] has shown near-software inference accuracy on the binary neural network, however it is known that the weight precision requirement of online training is higher than that of inference.”
BackgroundSelf-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
In-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
Chakraborty, I., Ali, M., Ankit, A., Jain, S., Roy, S., Sridharan, S., Agrawal, A., Raghunathan, A., & Roy, K. (2020) · Proceedings of the IEEE · doi:10.1109/jproc.2020.3003007 · Cited by 140
“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].”
BackgroundMethodInfluentialSelf-rectifying resistive memory in passive crossbar arrays
Jeon, K., Kim, J., Ryu, J. J., Yoo, S.-J., Song, C., Yang, M., Jeong, D. S., & Kim, G. H. (2021) · Nature Communications · doi:10.1038/s41467-021-23180-2 · Cited by 145
“As a result, in-memory processors based on non-volatile memory frequently employ RRAM loaded with weight matrices 12 – 16 .”
MethodExploring the Connection Between Binary and Spiking Neural Networks
Lu, S. K., & Sengupta, A. (2020) · Frontiers in Neuroscience · doi:10.3389/fnins.2020.00535 · Cited by 134
“…[8] already supports fixed point precision (as low as 4 bits), algorithmic progress on BNNs have contributed to the recent wave of specialized “In-Memory” BNN hardware accelerators using CMOS [9], [10] and post-CMOS technologies [11] that are highly optimized for single-bit state representations.”
BackgroundPeng, X., Liu, R., & Yu, S. (2019) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas.2019.8702715 · Cited by 131
“Table II summarizes the estimation results of the 8-bit RRAM architecture for all the convolutional layers in VGG-16 [10], this design could also extend to XNOR-RRAM [11] for binary neural network.”
MethodSelf-citationA Configurable Multi-Precision CNN Computing Framework Based on Single Bit RRAM
Zhu, Z., Sun, H., Lin, Y., Dai, G., Xia, L., Han, S., Wang, Y., & Yang, H. (2019) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1145/3316781.3317739 · Cited by 117
“In order to achieve high reliable and accurate CNN computing, researchers and chip designers also proposed several CNN accelerators based on single bit RRAM [2, 17, 18, 21].”
BackgroundPeng, X., Liu, R., & Yu, S. (2019) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2019.2958568 · Cited by 108
“According to the prior work on binary neural network where 3-bit ADC was reportedly sufficient for partial-sum-collecting in 128 × 128 sub-array [28], we analyzed the effect of ADC precision on 8-bit networks, as a result, the 5-bit ADC is suffi-cient to provide high accuracy ( ∼ 90”
Self-citationHirtzlin, T., Bocquet, M., Penkovsky, B., Klein, J.-O., Nowak, E., Vianello, E., Portal, J.-M., & Querlioz, D. (2020) · Frontiers in Neuroscience · doi:10.3389/fnins.2019.01383 · Cited by 93
Yin, S., Kim, Y., Han, X., Barnaby, H., Yu, S., Luo, Y., He, W., Sun, X., Kim, J., & Seo, J.-S. (2019) · IEEE Micro · doi:10.1109/mm.2019.2943047 · Cited by 101
Self-citationIn-Memory Computing in Emerging Memory Technologies for Machine Learning: An Overview
Roy, K., Chakraborty, I., Ali, M., Ankit, A., & Agrawal, A. (2020) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18072.2020.9218505 · Cited by 118
“However, due to crossbar device and circuit-level nonidealities (described in the next sub-section) and data-sparsity, recent proposals [20, 21] adopt lower-than-ideal ADC precision for better performance/energy.”
BackgroundHfO 2 -based resistive switching memory devices for neuromorphic computing
Brivio, S., Spiga, S., & Ielmini, D. (2022) · Neuromorphic Computing and Engineering · doi:10.1088/2634-4386/ac9012 · Cited by 119
“LRS and HRS) in binarized [97, 98] and ternary neural networks (TNNs) [99].”
Compute-in-Memory with Emerging Nonvolatile-Memories: Challenges and Prospects
Yu, S., Sun, X., Peng, X., & Huang, S. (2020) · Custom Integrated Circuits Conference · doi:10.1109/cicc48029.2020.9075887 · Cited by 87
Self-citationFORMS: Fine-grained Polarized ReRAM-based In-situ Computation for Mixed-signal DNN Accelerator
Yuan, G., Behnam, P., Li, Z., Shafiee, A. R., Lin, S., Ma, X., Liu, H., Qian, X., Bojnordi, M. N., Wang, Y., et al. (2021) · Proceedings - International Symposium on Computer Architecture/Proceedings · doi:10.1109/isca52012.2021.00029 · Cited by 98
A Review of In-Memory Computing Architectures for Machine Learning Applications
Bavikadi, S., Sutradhar, P. R., Khasawneh, K. N., Ganguly, A., & Dinakarrao, S. M. P. (2020) · ACM Great Lakes Symposium on VLSI (GLSVLSI) · doi:10.1145/3386263.3407649 · Cited by 86
“XNOR-SRAM: XNOR-SRAM[26]isanin-memorybinary/ternary convolutionmacroforacceleratingneuralnetworks.”
BackgroundMethodInfluentialSTT-BNN: A Novel STT-MRAM In-Memory Computing Macro for Binary Neural Networks
Pham, T.-N., Trinh, Q., Chang, I., & Alioto, M. (2022) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2022.3169759 · Cited by 77
A Robust 8-Bit Non-Volatile Computing-in-Memory Core for Low-Power Parallel MAC Operations
Zhang, S., Huang, K., & Shen, H. (2020) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2020.2971642 · Cited by 64
“As a result, it achieves a better energy efficiency than the inverting amplifier based scheme [21], [36]–[38].”
BackgroundPXNOR-BNN: In/With Spin-Orbit Torque MRAM Preset-XNOR Operation-Based Binary Neural Networks
Chang, L., Ma, X., Wang, Z., Zhang, Y., Xie, Y., & Zhao, W. S. (2019) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2019.2926984 · Cited by 63
“2) BNNs: On the image classification benchmarks such as CIFAR-10, BNNs have been demonstrated to achieve nearly state-of-the-art accuracy [9], [23].”
BackgroundTiM-DNN: Ternary In-Memory Accelerator for Deep Neural Networks
Jain, S., Gupta, S. K., & Raghunathan, A. (2020) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2020.2993045 · Cited by 66
“In this section, we limit our discussion to in-memory computing for DNNs [20]–[30], [49]–[53].”
BackgroundMethodInfluentialPark, J., Kim, S., Song, M. S., Youn, S., Kim, K., Kim, T.-H., & Kim, H. (2024) · ACS Applied Materials & Interfaces · doi:10.1021/acsami.3c13775 · Cited by 56
CIMAT: A Compute-In-Memory Architecture for On-chip Training Based on Transpose SRAM Arrays
Jiang, H., Peng, X., Huang, S., & Yu, S. (2020) · IEEE Transactions on Computers · doi:10.1109/tc.2020.2980533 · Cited by 69
“As a trade-off, the limited precision of ADCs and their variations lead to approximate computation results in CIM, which generally result in a slight degradation of the inference accuracy [19].”
BackgroundSelf-citationPeng, H., Gan, L., & Guo, X. (2024) · Chip · doi:10.1016/j.chip.2024.100093 · Cited by 69
Giacomin, E., Greenberg-Toledo, T., Kvatinsky, S., & Gaillardon, P. (2018) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2018.2872455 · Cited by 53
“To further reduce the overall energy consumption of such systems, some recent works [11]–[15] on Binary Neural Networks (BNNs) have opened the paths to more efficient convolutional blocks.”
MethodResultOutstanding Bit Error Tolerance of Resistive RAM-Based Binarized Neural Networks
Hirtzlin, T., Bocquet, M., Klein, J.-O., Nowak, E., Vianello, E., Portal, J.-M., & Querlioz, D. (2019) · 2019 IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas.2019.8771544 · Cited by 51
Lee, J.-H., Lim, D., Jeong, H., Ma, H., & Shi, L. (2019) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2019.2906249 · Cited by 58
“Many works [10], [17], [18] have implemented inference chips using famous QNN models, such as BNN [25] or XNORNet [26].”
MethodReconfigurable Field-Effect Transistor as a Synaptic Device for XNOR Binary Neural Network
Bae, J., Kim, H., Kwon, D., Lim, S., Lee, S., Park, B., & Lee, J. (2019) · IEEE Electron Device Letters · doi:10.1109/led.2019.2898448 · Cited by 61
“Therefore, binary neural networks (BNNs) have been studied to simplify neural network and to efficiently perform computation in neural network by reducing the complexity of the operations and the number of parameters [4]–[11].”
BackgroundMethodInfluentialStochastic Computing for Hardware Implementation of Binarized Neural Networks
Hirtzlin, T., Penkovsky, B., Bocquet, M., Klein, J., Portal, J., & Querlioz, D. (2019) · IEEE Access · doi:10.1109/access.2019.2921104 · Cited by 57
Woo, J., & Yu, S. (2019) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2019.2917764 · Cited by 52
“3 bits from the obtained current range should be needed to achieve a pattern recognition accuracy > 98% for the MNIST dataset [39], but the reduced I Real makes it difficult to have enough weighted sum precision.”
BackgroundSelf-citationIn-Memory Computing for Machine Learning and Deep Learning
Lepri, N., Glukhov, A., Cattaneo, L., Farronato, M., Mannocci, P., & Ielmini, D. (2023) · IEEE Journal of the Electron Devices Society · doi:10.1109/jeds.2023.3265875 · Cited by 52
“Depending on the BNN, the resulting current sum can also be directly compared to a reference current by means of a sense amplifier [138], thus operating a threshold-type activation function.”
BackgroundMethodMask Technique for Fast and Efficient Training of Binary Resistive Crossbar Arrays
Fouda, M. E., Lee, S., Lee, J., Eltawil, A., & Kurdahi, F. (2019) · IEEE Transactions on Nanotechnology · doi:10.1109/tnano.2019.2927493 · Cited by 41
“[16]–[18], the authors use 1T1R crossbar structure and proposed two approaches; 1) sequential BNNs where the current sensing circuit (neuron) is shared between all the neurons so only one column is activated at a time and 2) parallel BNNs where a current sensing circuit is assigned for each neuron.”
Qu, S., Li, B., Wang, Y., Xu, D., Zhao, X., & Zhang, L. (2020) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18072.2020.9218724 · Cited by 45
“However , their solutions either entail restructuring RRAM array [18, 19] or are subject to certain network architectures [9].”
BackgroundWu, J., Mo, F., Saraya, T., Hiramoto, T., & Kobayashi, M. (2020) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2020.3033831 · Cited by 42
Computing-in-Memory with SRAM and RRAM for Binary Neural Networks
Sun, X., Liu, R., Peng, X., & Yu, S. (2018) · IEEE International Conference on Solid-State and Integrated Circuit Technology · doi:10.1109/icsict.2018.8565811 · Cited by 47
“The estimated energy-efficiency of XNOR-SRAM and XNOR-RRAM shows ~30X improvement compared to the corresponding conventional SRAM and RRAM architectures with sequential row-by-row read-out.”
MethodInfluentialSelf-citationQiu, K., Jao, N., Zhao, M., Mishra, C. S., Akbulut, G. G., Jose, S., Sampson, J., Kandemir, M., & Narayanan, V. (2020) · IEEE International Symposium on High-Performance Computer Architecture (HPCA) · doi:10.1109/hpca47549.2020.00034 · Cited by 50
“The impact of different precisions on CNN accuracy has been extensively studied [ 24], [25], [26], [27], [28], [29], [30], [31], [32], [33].”
BackgroundCompact Single-Phase-Search Multistate Content-Addressable Memory Design Using One FeFET/Cell
Rajaei, R., Sharifi, M. M., Kazemi, A., Niemier, M. T., & Hu, X. S. (2020) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2020.3039477 · Cited by 40
“However, some promising solutions, such as the write-verify scheme [32], [33], were already proposed to cope with variations.”
BackgroundTutorial on memristor-based computing for smart edge applications
Gebregiorgis, A., Singh, A., Yousefzadeh, A., Wouters, D. J., Bishnoi, R., Catthoor, F., & Hamdioui, S. (2023) · Memories - Materials Devices Circuits and Systems · doi:10.1016/j.memori.2023.100025 · Cited by 38
Zheng, Q., Wang, Z., Feng, Z., Yan, B., Cai, Y., Huang, R., Chen, Y., Yang, C.-L., & Li, H. H. (2020) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18072.2020.9218590 · Cited by 46
“of accelerating DCNN executions [1]–[5].”
Shreya, S., Verma, G., Piramanayagam, S. N., & Kaushik, B. K. (2020) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2020.3038140 · Cited by 35
“…thus, could not explain the activation of the neuron based on the XAA module for all-spin NN. Similarly, resistive random access memory (RRAM) and pulse-code modulation (PCM)-based BNN designs (with CiM) require more operational voltage and involve more digital peripheral circuits [2], [5].”
BackgroundMethodInfluentialAn ADC-Less RRAM-Based Computing-in-Memory Macro With Binary CNN for Efficient Edge AI
Li, Y., Chen, J., Wang, L., Zhang, W., Guo, Z., Wang, J., Han, Y., Li, Z., Wang, F., Dou, C., et al. (2023) · IEEE Transactions on Circuits and Systems II: Express Briefs · doi:10.1109/tcsii.2022.3233396 · Cited by 38
“11 (right), using the HLMC 40 nm process, the energy cost of inverters dramatically decreases from 87% to 29% owing to the drop of Inv from 1.8V to 0.9V, achieving a comparable projected energy efficiency of 289.1 TOPS/W [13], [14], [15].”
MethodInfluentialEnabling Lower-Power Charge-Domain Nonvolatile In-Memory Computing With Ferroelectric FETs
Yin, G., Cai, Y., Wu, J., Duan, Z., Zhu, Z., Liu, Y., Wang, Y., Yang, H., & Li, X. (2021) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2021.3049844 · Cited by 37
Sutradhar, P. R., Bavikadi, S., Connolly, M., Prajapati, S., Indovina, M. A., Dinakarrao, S. M. P., & Ganguly, A. (2021) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/tpds.2021.3066909 · Cited by 52
“Therefore, it is common to see the PIM designers opting for lower data-precision such as binary or ternary weighted NNs [3], [10], [17], [18], [19], [20], or even binary NNs [6], [17], [19], [25], [26], [27].”
BackgroundHigh-Performance and Robust Spintronic/CNTFET-Based Binarized Neural Network Hardware Accelerator
Nasab, M. T., Amirany, A., Moaiyeri, M. H., & Jafari, K. (2022) · IEEE Transactions on Emerging Topics in Computing · doi:10.1109/tetc.2022.3202113 · Cited by 33
“However, in the proposed design and the designs presented in [5], [13], the learned weights are written on MTJs once after the learning process.”
BackgroundMethodInfluentialThe viability of analog-based accelerators for neuromorphic computing: a survey
Musisi‐Nkambwe, M., Afshari, S., Barnaby, H., Kozicki, M. N., & Esqueda, I. S. (2021) · Neuromorphic Computing and Engineering · doi:10.1088/2634-4386/ac0242 · Cited by 36
“Another is reduced precision, some studies have shown higher tolerances to accuracy degradation when ADC precision is reduced [116].”
BackgroundA Binarized Neural Network Approach to Accelerate in-Vehicle Network Intrusion Detection
Zhang, L., Yan, X., & Ma, D. (2022) · IEEE Access · doi:10.1109/access.2022.3208091 · Cited by 44
“This requirement for memory accesses is often the bottleneck of system performance as well as energy efficiency [22].”
BackgroundKwon, D., Park, E. C., Shin, W., Koo, R., Hwang, J., Bae, J., Kwon, D., & Lee, J. (2023) · Advanced Intelligent Systems · doi:10.1002/aisy.202300125 · Cited by 32
MAX2: An ReRAM-Based Neural Network Accelerator That Maximizes Data Reuse and Area Utilization
Mao, M., Peng, X., Liu, R., Li, J., Yu, S., & Chakrabarti, C. (2019) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2019.2908937 · Cited by 27
“Nonlinear quantization has been shown to achieve better accuracy than linear quantization for the same number of quantization levels [22].”
BackgroundMethodInfluentialSelf-citationScherer, M., Rutishauser, G., Cavigelli, L., & Benini, L. (2021) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2021.3075420 · Cited by 31
MINT: Mixed-Precision RRAM-Based IN-Memory Training Architecture
Jiang, H., Huang, S., Peng, X., & Yu, S. (2020) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas45731.2020.9181020 · Cited by 26
Self-citationSTT-BSNN: An In-Memory Deep Binary Spiking Neural Network Based on STT-MRAM
Nguyen, V.-T., Trinh, Q., Zhang, R., & Nakashima, Y. (2021) · IEEE Access · doi:10.1109/access.2021.3125685 · Cited by 25
“The proposed IMC designs in [8], [9] employ resistive RAM (RRAM), yielding low standby power and improving array density.”
MethodIn-memory batch-normalization for resistive memory based binary neural network hardware
Kim, H., Kim, Y., & Kim, J. (2019) · Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1145/3287624.3287718 · Cited by 27
“Therefore, several approaches to implement RCA-based BNN accelerators have been proposed recently [3, 4, 5, 6].”
MethodBenchmark of Ferroelectric Transistor-Based Hybrid Precision Synapse for Neural Network Accelerator
Luo, Y., Wang, P., Peng, X., Sun, X., & Yu, S. (2019) · IEEE Journal on Exploratory Solid-State Computational Devices and Circuits · doi:10.1109/jxcdc.2019.2925061 · Cited by 25
“To meet the requirement for data-centric computation, a new computing paradigm, compute-in-memory (CIM), has been proposed [7]–[10].”
BackgroundMethodSelf-citationLee, S., Kwon, D., Kim, H., Yoo, H.-N., & Lee, J. (2020) · IEEE Access · doi:10.1109/access.2020.3004045 · Cited by 23
“In recent neuromorphic research, emerging devices such as phase change memory (PCM) [33], and RRAM [8], [19] have been often investigated as synaptic devices.”
BackgroundInfluentialQuarry: Quantization-based ADC Reduction for ReRAM-based Deep Neural Network Accelerators
Azamat, A., Asim, F., & Lee, J. (2021) · International Conference on Computer Aided Design · doi:10.1109/iccad51958.2021.9643502 · Cited by 26
Memristor-based adaptive analog-to-digital conversion for efficient and accurate compute-in-memory
Hong, H., Du, Z., Jiang, M., Mao, R., Ren, Y., Li, F., Mao, W., Peng, M., Zhang, W., Liu, Z., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-65233-w · Cited by 34
Global-Gate Controlled One-Transistor One-Digital-Memristor Structure for Low-Bit Neural Network
Huang, M., Zhao, G., Wang, X., Zhang, W., Coquet, P., Tay, B. K., Zhong, G., & Li, J. (2020) · IEEE Electron Device Letters · doi:10.1109/led.2020.3037203 · Cited by 21
SiO 2 Fin-Based Flash Synaptic Cells in AND Array Architecture for Binary Neural Networks
Lee, S., Kim, H., Lee, S., Park, B., & Lee, J. (2021) · IEEE Electron Device Letters · doi:10.1109/led.2021.3125966 · Cited by 22
Hybrid MTJ/CNTFET-Based Binary Synapse and Neuron for Process-in-Memory Architecture
Nasab, M. T., Amirany, A., Moaiyeri, M. H., & Jafari, K. (2023) · IEEE Magnetics Letters · doi:10.1109/lmag.2023.3238271 · Cited by 21
Bai, J., Sun, S., Zhao, W. S., & Kang, W. (2023) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2023.3298705 · Cited by 28
“However, high-resolution ADCs introduce high-energy consumption, latency, and area requirements [23], [24].”
BackgroundZhao, C., Fang, J., Jiang, J., Xue, X., & Zeng, X. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2022.3215535 · Cited by 20
“Moreover, the partial sums need to be quantified frequently by high-resolution ADCs to reduce the quantization errors, inevitably increasing the hardware costs further [13].”
BackgroundTernary Output Binary Neural Network With Zero-Skipping for MRAM-Based Digital In-Memory Computing
Na, T. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3241446 · Cited by 19
“However, most state-of-the-art works have used two bit-cells for one synapse, storing one weight value to implement the BNN’s XNOR operation [15], [16], [17], [18], resulting in substantial area overheads.”
BackgroundImplementing Binarized Neural Networks with Magnetoresistive RAM without Error Correction
Hirtzlin, T., Penkovsky, B., Klein, J., Locatelli, N., Vincent, A. F., Bocquet, M., Portal, J., & Querlioz, D. (2019) · IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH) · doi:10.1109/nanoarch47378.2019.181300 · Cited by 24
Kim, H., Kim, Y., Ryu, S., & Kim, J. (2020) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18072.2020.9218657 · Cited by 23
“To address the limited array size issue, techniques such as partial sum quantization method using low-precision ADCs [18] or input-splitting method [19] have been proposed.”
BackgroundMethodNeural Network-Hardware Co-design for Scalable RRAM-based BNN AcceleratorsPreprint
Kim, Y., Kim, H., & Kim, J. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.02187 · Cited by 26
“To handle the issue, BCNN-RRAM [7] and XNOR-RRAM [8] quantized the partial sum and reduced the resolution of the ADC to 4 bits and 3 bits, respectively.”
BackgroundMethodResultInfluentialEnergy-efficient XNOR-free in-memory BNN accelerator with input distribution regularization
Kim, H., Oh, H., & Kim, J. (2020) · doi:10.1145/3400302.3415641 · Cited by 20
“Due to the great synergy of in-memory computing and BNNs, several in-memory BNN accelerators have been proposed [10, 14, 17, 19, 21].”
BackgroundPartial-Sum Quantization for Near ADC-Less Compute-In-Memory Accelerators
Saxena, U., & Roy, K. (2023) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1109/islped58423.2023.10244291 · Cited by 20
A Configurable BNN ASIC using a Network of Programmable Threshold Logic Standard Cells
Wagle, A., Khatri, S. P., & Vrudhula, S. B. K. (2020) · IEEE International Conference on Computer Design: VLSI in Computers and Processors (ICCD) · doi:10.1109/iccd50377.2020.00079 · Cited by 18
Implementation of Ternary Weights With Resistive RAM Using a Single Sense Operation Per Synapse
Laborieux, A., Bocquet, M., Hirtzlin, T., Klein, J., Nowak, E., Vianello, E., Portal, J., & Querlioz, D. (2020) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2020.3031627 · Cited by 16
Merged Logic and Memory Fabrics for Accelerating Machine Learning Workloads
Crafton, B., Spetalnick, S. D., Fang, Y., & Raychowdhury, A. (2020) · IEEE Design and Test · doi:10.1109/mdat.2020.3016587 · Cited by 15
“Originally used in FPGAs to take advantage of their configurability and offset the area and power overhead, BNNs have also been shown to work with RRAM in both simulation [116] and experiment [23].”
MethodJiang, H., Peng, X., Huang, S., & Yu, S. (2019) · International Symposium on Memory Systems (MEMSYS) · doi:10.1145/3357526.3357552 · Cited by 18
“As a trade-off, the limited precision of ADCs and their variations lead to approximate computation results in CIM, which generally result in a slight degradation of the inference accuracy [12].”
BackgroundSelf-citationNevarez, Y., Rotermund, D., Pawelzik, K. R., & García-Ortiz, A. (2021) · IEEE Access · doi:10.1109/access.2021.3085216 · Cited by 16
“In [36], Sun et al. report an accuracy of 98 .”
BackgroundAn encoding framework for binarized images using hyperdimensional computing
Smets, L., Leekwijck, W. V., Tsang, I. J., & Latré, S. (2024) · Frontiers in Big Data · doi:10.3389/fdata.2024.1371518 · Cited by 22
Ultrafast and Energy-Efficient Ferrimagnetic XNOR Logic Gates for Binary Neural Networks
Wang, G., Zhang, Y., Zhang, Z., Zheng, Z., Zhang, K., Wang, J., Klein, J., Ravelosona, D., & Zhao, W. S. (2021) · IEEE Electron Device Letters · doi:10.1109/led.2021.3062382 · Cited by 15
“Based on the emerging non-volatile memories (NVMs), such as resistive random-access memory (RRAM) [6]–[8], magnetic RAM (MRAM), [9]–[11] and domain wall (DW) based race-track memory [12]–[14], BNNs can further take advantages of in-memory-computing (IMC) architecture by performing bit-operations…”
BackgroundMulti-Input Logic-in-Memory for Ultra-Low Power Non-Von Neumann Computing
Zanotti, T., Pavan, P., & Puglisi, F. M. (2021) · Micromachines · doi:10.3390/mi12101243 · Cited by 16
“While the in-memory computation of the analog vector matrix multiplication is extremely attractive [6,7,50,52], it has some limitations and tradeoffs.”
BackgroundDevice quantization policy in variation-aware in-memory computing design
Chang, C.-C., Li, S.-T., Pan, T.-L., Tsai, C.-M., Wang, I., Chang, T., & Hou, T. (2022) · Scientific Reports · doi:10.1038/s41598-021-04159-x · Cited by 16
Ahmed, S. T., & Tahoori, M. B. (2022) · IEEE International Test Conference (ITC) · doi:10.1109/itc50671.2022.00032 · Cited by 14
“We have used the weight matrix splitting and current accumulation as proposed in [27].”
MethodWei, W.-C., Jhang, C.-J., Chen, Y., Xue, C.-X., Sie, S.-H., Lee, J.-L., Kuo, H.-W., Lu, C.-C., Chang, M., & Tang, K. (2020) · IEEE Journal on Exploratory Solid-State Computational Devices and Circuits · doi:10.1109/jxcdc.2020.2992306 · Cited by 14
“Sun et al. [8] described how to optimize the accuracy and chip area when mapping the weights to different crossbar sizes and different ADC precisions; however, these works did not consider partial sum quantization during training, which may cause accuracy loss in silicon realization.”
BackgroundHigh Performance and Low Power Spintronic Binarized Neural Network Hardware Accelerator
Nasab, M. T., Amirany, A., Moaiyeri, M. H., & Jafari, K. (2022) · Iranian Conference on Electrical Engineering · doi:10.1109/icee55646.2022.9827189 · Cited by 14
Quantized convolutional neural networks: a hardware perspective
Zhang, L., Krestinskaya, O., Fouda, M. E., Eltawil, A. M., & Salama, K. N. (2025) · Frontiers in Electronics · doi:10.3389/felec.2025.1469802 · Cited by 18
Mapping Binary ResNets on Computing-In-Memory Hardware with Low-bit ADCs
Kim, Y., Kim, H., Park, J., Oh, H., & Kim, J. (2021) · Design, Automation, and Test in Europe · doi:10.23919/date51398.2021.9474056 · Cited by 13
Radiation-Immune Spintronic Binary Synapse and Neuron for Process-in-Memory Architecture
Nasab, M. T., Amirany, A., Moaiyeri, M. H., & Jafari, K. (2024) · IEEE Magnetics Letters · doi:10.1109/lmag.2024.3356815 · Cited by 13
CIM-MLC: A Multi-level Compilation Stack for Computing-In-Memory Accelerators
Qu, S., Zhao, S., Li, B., He, Y., Cai, X., Zhang, L., & Wang, Y. (2024) · International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · doi:10.1145/3620665.3640359 · Cited by 18
“…[39] ◆ Jia et al. [29] PUMA [4] IMDP [19] FPSA [28] ◆ Neural Cache [18] ◆ ConvSRAM [6] Zhu et al. [51] SRE [47] nvCIM [46] XNOR-RRAM [43] ◆ XNOR-SRAM [49] ⚫ Han et al. [23] MAX2 [34] ReRAM-RNN [33] FORMS [50] ⚫ GUO et al. [21] ReRAM-Based ◆ SRAM-Based ⚫ FLASH-Based ◆ Vesti…”
BackgroundMethodOvercoming Challenges for Achieving High in-situ Training Accuracy with Emerging Memories
Huang, S., Sun, X., Peng, X., Jiang, H., & Yu, S. (2020) · Design, Automation, and Test in Europe · doi:10.23919/date48585.2020.9116215 · Cited by 19
Self-citationSTT-MRAM Architecture with Parallel Accumulator for In-Memory Binary Neural Networks
Pham, T.-N., Trinh, Q., Chang, I., & Alioto, M. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401695 · Cited by 13
Processing-in-Memory Technology for Machine Learning: From Basic to ASIC
Taylor, B., Zheng, Q., Li, Z., Li, S., & Chen, Y. (2022) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2022.3168404 · Cited by 14
Accelerating In-Vehicle Network Intrusion Detection System Using Binarized Neural Network
Zhang, L., Yan, X., & Ma, D. (2022) · SAE International Journal of Advances and Current Practices in Mobility · doi:10.4271/2022-01-0156 · Cited by 13
Zhang, Y., Wang, X., Jiang, X., Yang, Y., Shen, Z., & Jia, Z. (2022) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2022.102531 · Cited by 13
ReBNN: in-situ acceleration of binarized neural networks in ReRAM using complementary resistive cell
Song, L., Wu, Y., Qian, X., Li, H. H., & Chen, Y. (2019) · CCF Transactions on High Performance Computing · doi:10.1007/s42514-019-00014-8 · Cited by 13
“…based accelerators for neural networks are (Chen and Li 2018a; Chen et al. 2018b, 2019a, b; Cheng et al. 2017; Ji et al. 2018a, 2019; Li et al. 2018; Lin et al. 2019a, b; Liu et al. 2018c, 2015; Mao et al. 2019, 2018a; Qiao et al. 2018; Sun et al. 2018a, b; Tang et al. 2017; Wang et al. 2018).”
BackgroundMethodDot‐Product Operation in Crossbar Array Using a Self‐Rectifying Resistive Device
Jeon, K., Ryu, J. J., Jeong, D. S., & Kim, G. H. (2022) · Advanced Materials Interfaces · doi:10.1002/admi.202200392 · Cited by 15
Metadata
- Authors
- Sun, X., Yin, S., Peng, X., Liu, R., Seo, J.-S., & Yu, S.
- Affiliations
- Arizona State University (US)
- Venue
- 2018 Design, Automation & Test in Europe Conference & Exhibition (DATE)
- Publisher
- IEEE
- Published
- 2018-03
- Pages
- 1423-1428
- OpenAlex
- W2798554798
- Cited by
- 246
- References
- 20
- Citations by year
- 2026: 7 · 2025: 17 · 2024: 33 · 2023: 34 · 2022: 40 · 2021: 35 · 2020: 42 · 2019: 30 · 2018: 8
- FWCI
- 45.74
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Design, Automation, and Test in Europe (publishedVersion)
- Funders
- Canadian Institute for Advanced Research
- 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 › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- resistive random access memory, binary neural networks, synaptic structure, array partitioning, energy-efficient neuromorphic hardware, convolutional neural networks, fully connected layers, MNIST, CIFAR-10
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.