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].”
Method
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 192
Hänsch, W., Raghunathan, A., Roy, K., Chakrabarti, B., Phatak, C., Wang, C., & Guha, S. (2023) · 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 157
“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
X-Former: In-Memory Acceleration of Transformers
Sridharan, S., Stevens, J. R., Roy, K., & Raghunathan, A. (2023) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2023.3282046 · Cited by 68
“and energy but also affects the lifetime of the NVM devices, since they have very limited endurance [8], [9], [10].”
BackgroundSelf-citationCompute-in-Memory Technologies and Architectures for Deep Learning Workloads
Ali, M., Roy, S., Saxena, U., Sharma, T., Raghunathan, A., & Roy, K. (2022) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2022.3203583 · Cited by 66
“In this article, we use the term CiM to denote compute operations performed within a memory unit, including static random access memory (SRAM), DRAM, and/or nonvolatile memory (NVM) technologies [5], [6].”
MethodSelf-citationA Review of Graphene‐Based Memristive Neuromorphic Devices and Circuits
Walters, B., Jacob, M. V., Amirsoleimani, A., & Azghadi, M. R. (2023) · Advanced Intelligent Systems · doi:10.1002/aisy.202300136 · Cited by 59
Lead-Free Cs 3 Bi 2 I 9 Perovskite Memristors for Energy-Efficient Neuromorphic Computing
Vishwanath, S. K., Karkera, C., Mohammad, T., Sharma, P., Badavathu, R. N., Khandelwal, U., Kanwat, A., Chakrabarty, P., Suresh, D. I., Sahay, S., et al. (2025) · ACS Energy Letters · doi:10.1021/acsenergylett.5c00411 · Cited by 55
Circuits and Architectures for In-Memory Computing-Based Machine Learning Accelerators
Ankit, A., Chakraborty, I., Agrawal, A., Ali, M., & Roy, K. (2020) · IEEE Micro · doi:10.1109/mm.2020.3025863 · Cited by 46
Self-citationYadav, R. S., Gupta, P., Holla, A., Khan, K. I. A., Muduli, P. K., & Bhowmik, D. (2023) · ACS Applied Electronic Materials · doi:10.1021/acsaelm.2c01488 · Cited by 40
Safari, M. M., & Pourrostam, J. (2024) · Signal Processing · doi:10.1016/j.sigpro.2024.109446 · Cited by 36
The 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
“…combination of memory devices to extend the conductance range and improve linearity of weight updates could also be considered, for example suggested in [63] is a PCM for MSB, and transistors to cap for LSB thus relieving training on PCM due to its high write latencies and resistance drifting.”
BackgroundInfluentialMagnetoresistive Circuits and Systems: Embedded Non-Volatile Memory to Crossbar Arrays
Agrawal, A., Wang, C., Sharma, T., & Roy, K. (2021) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3069682 · Cited by 33
“Although such architectures present excellent energy-efficiency, the analog nature of computing introduces errors and approximations, which limits their use to error-tolerant applications only [139].”
BackgroundSelf-citationKwon, 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 34
Kim, J., Im, J., Shin, W., Lee, S., Oh, S., Kwon, D., Jung, G., Choi, W. Y., & Lee, J. (2024) · Advanced Science · doi:10.1002/advs.202308460 · Cited by 28
Kwon, D., Woo, S. Y., Lee, K.-H., Hwang, J., Kim, H., Park, S., Shin, W., Bae, J., Kim, J., & Lee, J. (2023) · Science Advances · doi:10.1126/sciadv.adg9123 · Cited by 29
“However, the increasing area and power overhead of von Neumann computing architectures, mainly induced by the data communication between processing units and memory, have also become very critical issues in DNNs ( 10 – 14 ).”
Krestinskaya, O., Zhang, L., & Salama, K. N. (2023) · IEEE Transactions on Nanotechnology · doi:10.1109/tnano.2023.3293026 · Cited by 29
“ADC necessary resolution is calculated as ceil(log2((2DAC − 1)× (2w − 1)× i)), where bDAC is a DAC resolution, i is a number of crossbar rows, and bw is the number of bits in weights [9].”
BackgroundMethodInfluentialAn Analysis of Components and Enhancement Strategies for Advancing Memristive Neural Networks
Park, H., Han, J., Yim, S., Shin, D. H., Park, T. W., Woo, K. S., Lee, S. H., Cho, J. M., Kim, H. W., Park, T., et al. (2025) · Advanced Materials · doi:10.1002/adma.202412549 · Cited by 27
IMAGIN: Library of IMPLY and MAGIC NOR-Based Approximate Adders for In-Memory Computing
Jha, C. K., Thangkhiew, P. L., Datta, K., & Drechsler, R. (2022) · IEEE Journal on Exploratory Solid-State Computational Devices and Circuits · doi:10.1109/jxcdc.2022.3222015 · Cited by 29
“While approximate adders using memristor crossbars have been studied using analog computation [15], [16], the same is not true for digital computation.”
MethodRoy, K., Wang, C., Roy, S., Raghunathan, A., Yang, K., & Sengupta, A. (2024) · Nature Reviews Electrical Engineering · doi:10.1038/s44287-024-00107-9 · Cited by 38
“…that emerging neuroinspired architectures such as crossbar in-memory computing still face numerous challenges in processing large-scale AI tasks 157 , co-design of spintronic device and memory circuits for efficient digital architectures might serve as a near-term solution for accelerating…”
Self-citationAll‐Ferroelectric Spiking Neural Networks via Morphotropic Phase Boundary Neurons
Kim, J., Park, E. C., Shin, W., Koo, R., Im, J., Han, C., Lee, J., & Kwon, D. (2024) · Advanced Science · doi:10.1002/advs.202407870 · Cited by 25
Investigation of Low-Frequency Noise Characteristics in Gated Schottky Diodes
Kwon, D., Shin, W., Bae, J., Lim, S., Park, B., & Lee, J. (2021) · IEEE Electron Device Letters · doi:10.1109/led.2021.3051197 · Cited by 23
“RECENTLY, artificial neural networks (ANNs) using backpropagation algorithm have shown state-of-the-art performance in various fields such as speech recognition or image processing [1], [2].”
MethodOn-chip learning of a domain-wall-synapse-crossbar-array-based convolutional neural network
Desai, V. B., Kaushik, D., Sharda, J., & Bhowmik, D. (2022) · Neuromorphic Computing and Engineering · doi:10.1088/2634-4386/ac62db · Cited by 23
“On-chip learning is known to provide several advantages for edge devices in terms of data security and the like [6, 7].”
BackgroundPark, S., Seong, S., Jeon, G., Ji, W., Noh, K., Kim, S., & Chung, Y. (2022) · Advanced Electronic Materials · doi:10.1002/aelm.202200554 · Cited by 24
A Multilayer Neural Accelerator With Binary Activations Based on Phase-Change Memory
Bertuletti, M., Muñoz-Martín, I., Bianchi, S., Bonfanti, A. G., & Ielmini, D. (2023) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2022.3233292 · Cited by 22
“Note that an ADC consumes more than 80% and 60% of the circuit power and area, respectively [23], [24].”
BackgroundPIMCOMP: A Universal Compilation Framework for Crossbar-based PIM DNN Accelerators
Sun, X., Wang, X., Li, W., Wang, L., Han, Y., & Chen, X. (2023) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac56929.2023.10247928 · Cited by 27
“The 2D crossbar structure formed by NVM devices has attracted growing interest due to its high memory density and parallel in-situ computing properties [4].”
BackgroundCurrent opinions on memristor-accelerated machine learning hardware
Jiang, M., Xu, Y., Li, Z., & Li, C. (2025) · Current Opinion in Solid State and Materials Science · doi:10.1016/j.cossms.2025.101226 · Cited by 28
Enabling Robust SOT-MTJ Crossbars for Machine Learning using Sparsity-Aware Device-Circuit Co-design
Sharma, T., Wang, C., Agrawal, A., & Roy, K. (2021) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1109/islped52811.2021.9502492 · Cited by 22
Self-citationZanotti, T., Pavan, P., & Puglisi, F. M. (2022) · Microelectronic Engineering · doi:10.1016/j.mee.2022.111886 · Cited by 20
SAMBA: S parsity A ware In- M emory Computing B ased Machine Learning A ccelerator
Kim, D. E., Ankit, A., Wang, C., & Roy, K. (2023) · IEEE Transactions on Computers · doi:10.1109/tc.2023.3257513 · Cited by 17
Self-citationLarge-scale crossbar arrays based on three-terminal MoS2 memtransistors
Schranghamer, T. F., Pannone, A., Kumar, J. M., Baiju, D. K. T., Chen, C., Knight, T. V. M., Tadekawa, S., Haines, E., Ordoñez, R. C., Hayashi, C. K., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-64536-2 · Cited by 27
“Furthermore, as shown in Supplementary Note 10, experimental testing has veri fi ed that each gate access line is suf fi ciently electrostatically isolated to prevent programming of devices in adjacent rows, thus allowing for subsets of devices to be initialized without affecting the entire array.”
InfluentialVariability-Aware Training and Self-Tuning of Highly Quantized DNNs for Analog PIM
Deng, Z., & Orshansky, M. E. (2022) · Design, Automation, and Test in Europe · doi:10.23919/date54114.2022.9774625 · Cited by 21
Lee, K.-H., Kwon, D., Woo, S. Y., Ko, J., Choi, W. Y., Park, B., & Lee, J. (2022) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2022.3207707 · Cited by 17
“power increase and long latency due to the bottleneck caused by data transfer between the processing unit and the memory unit [3], [4], [5].”
BackgroundAdvancements in memory technologies for artificial synapses
Sehgal, A., Dhull, S., Roy, S., & Kaushik, B. K. (2024) · Journal of Materials Chemistry C · doi:10.1039/d3tc04131j · Cited by 17
Self-citationOn the Intrinsic Robustness of NVM Crossbars Against Adversarial Attacks
Roy, D., Chakraborty, I., Ibrayev, T., & Roy, K. (2021) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18074.2021.9586202 · Cited by 21
Self-citationHyperX: A Hybrid RRAM-SRAM partitioned system for error recovery in memristive Xbars
Kosta, A. K., Soufleri, E., Chakraborty, I., Agrawal, A., Ankit, A., & Roy, K. (2022) · Design, Automation, and Test in Europe · doi:10.23919/date54114.2022.9774549 · Cited by 16
Self-citationHe, K., Chakraborty, I., Wang, C., & Roy, K. (2022) · International Conference on Computer Aided Design · doi:10.1145/3508352.3549453 · Cited by 23
“Considering the ADC overhead and implication of crossbar non-idealities [4, 11, 23, 39], we choose array size of 64x64, 128x128, and 256x256.”
BackgroundSelf-citationYang, T. J., Cho, J. R., Lee, H., Lee, H. J., Myoung, S. J., Lee, D. Y., Choi, S., Bae, J., Kim, D. M., Kim, C., et al. (2024) · IEEE Access · doi:10.1109/access.2024.3366224 · Cited by 16
“Here, symmetry and linearity are the most crucial figure of merits (FoMs) of analog RS synapses [8], [9], [10], [11].”
BackgroundProspects and challenges of electrochemical random-access memory for deep-learning accelerators
Cui, J., Liu, H., & Cao, Q. (2024) · Current Opinion in Solid State and Materials Science · doi:10.1016/j.cossms.2024.101187 · Cited by 17
“…with the underlying peripheral silicon-CMOS circuits, especially the analog-to-digital and digital-to-analog converters as part of the controller interface which can account for substantial, i.e., over 50 % [27], of the total energy consumption and area of analog in-memory computing chips.”
Multi-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
Exploiting Non-idealities of Resistive Switching Memories for Efficient Machine Learning
Yon, V., Amirsoleimani, A., Alibart, F., Melko, R. G., Drouin, D., & Beilliard, Y. (2022) · Frontiers in Electronics · doi:10.3389/felec.2022.825077 · Cited by 18
“RSMs are subject to variability and performance issues (Adam et al., 2018; Wang et al., 2019a; Krestinskaya et al., 2019; Chakraborty et al., 2020; Zahoor et al., 2020; Zhao et al., 2020; Xi et al., 2021), which currently restrict their usage to small and noisy NN hardware implementations that can solve only simple pr…”
BackgroundBreaking the memory wall: next-generation artificial intelligence hardware
Roy, K., Kosta, A. K., Sharma, T., Negi, S., Sharma, D., Saxena, U., Roy, S., Raghunathan, A., Wan, Z., Spetalnick, S. D., et al. (2025) · Frontiers in Science · doi:10.3389/fsci.2025.1611658 · Cited by 19
Self-citationApproximate Computing and the Efficient Machine Learning Expedition
Henkel, J., Li, H., Raghunathan, A., Tahoori, M. B., Venkataramani, S., Yang, X., & Zervakis, G. (2022) · International Conference on Computer Aided Design · doi:10.1145/3508352.3561105 · Cited by 27
“In-memory computing is generally based on analog processing and has been explored in the context of SRAM [36] and non-voltatile memories such as ReRAM, MRAM, or PCM [8].”
MethodSelf-citationEnergy-Accuracy Trade-Offs for Resistive In-Memory Computing Architectures
Roy, S. K., & Shanbhag, N. R. (2024) · IEEE Journal on Exploratory Solid-State Computational Devices and Circuits · doi:10.1109/jxcdc.2024.3381888 · Cited by 16
“Recognizing this issue, there has been some work on analyzing the compute accuracy of resistive IMCs focusing primarily on resistive crossbars [26], [27], [28], [29], [30], [31], [32], [33], [34].”
BackgroundSelf-citationSySCIM: SystemC-AMS Simulation of Memristive Computation In-Memory
Shadmehri, S. H. H., BanaGozar, A., Kamal, M., Stuijk, S., Afzali‐Kusha, A., Pedram, M., & Corporaal, H. (2022) · Design, Automation, and Test in Europe · doi:10.23919/date54114.2022.9774749 · Cited by 16
Vahdat, S., Kamal, M., Afzali‐Kusha, A., & Pedram, M. (2021) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3105043 · Cited by 10
CODEX: Stochastic Encoding Method to Relax Resistive Crossbar Accelerator Design Requirements
Liu, T., Amirsoleimani, A., Xu, J., Alibart, F., Beilliard, Y., Ecoffey, S., Drouin, D., & Genov, R. (2022) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2022.3157789 · Cited by 13
Sahu, U., Sisodia, N., Sharda, J., Muduli, P. K., & Bhowmik, D. (2022) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2022.3142119 · Cited by 10
Metrics for spin-based computing
Kurebayashi, H., Finocchio, G., Everschor‐Sitte, K., Gartside, J. C., Taniguchi, T., Litvinenko, A. N., Kumar, A., Åkerman, J. J., Vasilaki, E., Selçuk, K., et al. (2026) · Nature Reviews Physics · doi:10.1038/s42254-025-00918-1 · Cited by 27
“One of the major challenges is to interconnect the neurons and synapses at high density, for which most studies 34 have focused on crossbar arrays of synapses to vary the synaptic weights through the electrical conductance of the devices.”
Smolka, M., Stoepel, L., Quill, J., Wahlbrink, T., Floehr, J., Böschen, S., Letmathe, P., & Lemme, M. C. (2024) · Springer eBooks · doi:10.1007/978-3-031-54700-3_10 · Cited by 10
A Marr's Three‐Level Analytical Framework for Neuromorphic Electronic Systems
Guo, Y., Zou, X., Hu, Y., Yang, Y., Wang, X., He, Y., Kong, R., Guo, Y., Li, G., Zhang, W., et al. (2021) · Advanced Intelligent Systems · doi:10.1002/aisy.202100054 · Cited by 16
Cui, Y., Li, J., Gu, C., Wang, C., & Liu, W. (2025) · IEEE Transactions on Nanotechnology · doi:10.1109/tnano.2025.3552433 · Cited by 7
“The dot-product operation is easy to achieve in RRAM arrays [9], [35], [36].”
Hsin, T., Lin, C., Wang, P., Yang, C., & Pai, C. (2025) · Advanced Science · doi:10.1002/advs.202417735 · Cited by 9
Power-Aware Quantization in Analog In-Memory Computing With STT-MRAM Macro
Zhou, M., Du, H., Guo, Y., & Cai, H. (2023) · IEEE Transactions on Magnetics · doi:10.1109/tmag.2023.3286467 · Cited by 8
“Over the past few years, numerous explorations of this emerging computing paradigm have been made to break the bottleneck of memory-processor communication of the traditional von Neumann architecture [1], [2], [3], [4].”
BackgroundTowards Efficient RRAM-based Quantized Neural Networks Hardware: State-of-the-art and Open Issues
Krestinskaya, O., Zhang, L., & Salama, K. N. (2022) · IEEE International Conference on Nanotechnology (IEEE-NANO) · doi:10.1109/nano54668.2022.9928590 · Cited by 7
“This can be solved by developing 3D integration techniques and improving lithography techniques for 3D stacking [1], which is also affected by selector device switching and non-linearity.”
BackgroundMethodInfluentialCho, K., Malhotra, A., & Gupta, S. K. (2023) · IEEE Journal on Exploratory Solid-State Computational Devices and Circuits · doi:10.1109/jxcdc.2023.3320677 · Cited by 7
Negi, S., Chakraborty, I., Ankit, A., & Roy, K. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.12125 · Cited by 7
“4 describes the basic structure of an MCA.”
BackgroundInfluentialSelf-citationKim, D. E., Sharma, T., & Roy, K. (2025) · IEEE transactions on circuits and systems for artificial intelligence. · doi:10.1109/tcasai.2025.3601975 · Cited by 10
“While emerging memories offer a high-density memory solution, they typically suffer from issues such as resistance drift, write endurance and/or low on-off distinguishability [50].”
Self-citationLiu, F., Yin, A., Xue, C., Wang, B., Feng, Z., Liu, H., Li, X., Gao, H., Xiong, T., & Si, X. (2024) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2024.3446045 · Cited by 6
“The power consumption overhead of the analog CIM circuit mainly comes from the quantization circuit [9].”
Yadav, R. S., Sadashiva, A., Holla, A., Muduli, P. K., & Bhowmik, D. (2023) · Neuromorphic Computing and Engineering · doi:10.1088/2634-4386/acf0e4 · Cited by 6
“and outer product operations in it, on-chip learning in crossbar arrays becomes very attractive for edge AI applications [4, 6].”
BackgroundCODEX: Stochastic Encoding Method to Relax Resistive Crossbar Accelerator Design RequirementsPreprint
Amirsoleimani, A., Liu, T., Xu, J., Alibart, F., Beilliard, Y., Ecoffey, S., Drouin, D., & Genov, R. (2021) · doi:10.36227/techrxiv.14485215.v2 · Cited by 6
“%) with analog-to-digital converters (ADCs) in particular making up over 57 % of peripheral circuit power in such platforms [12].”
BackgroundA Low-Power and Reliable RRAM-Based Configurable RO PUF With Aging Resilience
Li, J., Cui, Y., Wang, C., Gu, C., & Liu, W. (2025) · IEEE Transactions on Nanotechnology · doi:10.1109/tnano.2025.3569071 · Cited by 5
Kwon, D., Park, M., Kang, W.-M., Hwang, J., Koo, R., Bae, J., & Lee, J. (2023) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2023.3287824 · Cited by 6
“H ARDWARE-BASED neural networks (HNNs) utilizing analog synaptic devices with a nonvolatile memory function have been extensively studied as a solution to solve the problems of von Neumann computing architectures [1], [2], [3], [4], [5].”
MethodWang, K., Lü, J., Xiang, Z., Li, R., Wang, Z., Jin, H., & Jiang, R. (2024) · Applied Physics Letters · doi:10.1063/5.0212383 · Cited by 5
Function Approximation Using Analog Building Blocks in Flexible Electronics
Duarte, P. C. L., Dubé, A., Zervakis, G., Tahoori, M. B., & Nassif, S. (2025) · International Symposium on Quality Electronic Design (ISQED) · doi:10.1109/isqed65160.2025.11014355 · Cited by 6
“1 is a linear combination of B´ezier splines such that: While the basis function and multiplication can be implemented by multiply and accumulation using a crossbar array [20], the implementation of spline in analog domain is the focus.”
Hissariya, R., & Bhowmik, D. (2024) · IEEE Open Journal of Nanotechnology · doi:10.1109/ojnano.2024.3514900 · Cited by 4
StoX-Net: Stochastic Processing of Partial Sums for Efficient In-Memory Computing DNN Accelerators
Rogers, E. G., Mugdho, S. S., Gupte, K. K., & Wang, C. (2024) · IEEE International Conference on Rebooting Computing (ICRC) · doi:10.1109/icrc64395.2024.10937005 · Cited by 7
“Concretely, ADCs are found to consume over 60-80 % of the energy and chip area of IMC hardware [2], [6].”
Self-consistent Automated Parameter Extraction of RRAM Physics-Based Compact Model
Zanotti, T., Pavan, P., & Puglisi, F. (2022) · European Solid-State Device Research Conference · doi:10.1109/essderc55479.2022.9947161 · Cited by 4
“In fact, several works have demonstrated that RRAM devices enable the fabrication of in-memory computing hardware accelerators for deep learning applications [2], logic operations [3], but also the implementation of neuromorphic computing circuits which, by taking inspiration from the brain, promise to achieve extreme…”
BackgroundXu, J., Hong, H., Zhou, Y., Hu, X., & Duan, S. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.130517 · Cited by 3
Lambertini, A., Zanotti, T., Pavan, P., Padovani, A., & Puglisi, F. M. (2025) · APL Machine Learning · doi:10.1063/5.0274233 · Cited by 6
Lee, B., Ji, W., Kim, H., Han, S., Park, G., Hur, P., Jeon, G., Lee, H., Chung, Y., Son, J., et al. (2024) · Advanced Intelligent Systems · doi:10.1002/aisy.202400600 · Cited by 3
HEEPstor: an Open-Hardware Co-design Framework for Quantized Machine Learning at the Edge
Palacios, P., Medina, R., Ansaloni, G., & Atienza, D. (2025) · ACM International Conference on Computing Frontiers: Workshops and Special Sessions (CF Companion) · doi:10.1145/3706594.3726967 · Cited by 3
Yadav, R. S., Gupta, P., Yadav, V., Khan, K. I. A., Muduli, P. K., Sadashiva, A., & Bhowmik, D. (2025) · Physical Review Applied · doi:10.1103/22yc-b159 · Cited by 5
On Noise Stability and Robustness of Adversarially Trained Networks on NVM Crossbars
Tao, C., Roy, D., Chakraborty, I., & Roy, K. (2022) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2022.3193312 · Cited by 5
“Note, MVM operations are executed in the analog domain, and can significantly reduce power and latency when compared to digital von-Neumann CMOS architectures [13].”
MethodSelf-citationSpin Orbit Torque-based Crossbar Array for Error Resilient Binary Convolutional Neural Network
Danouchi, K., Prenat, G., & Anghel, L. (2022) · IEEE Latin American Test Symposium (LATS) · doi:10.1109/lats57337.2022.9936951 · Cited by 7
A memristive neural decoder for cryogenic fault-tolerant quantum error correction
Yon, V., Marcotte, F., Mouny, P.-A., Dagnew, G. A., Kulchytskyy, B., Rochette, S., Beilliard, Y., Drouin, D., & Ronagh, P. (2025) · Quantum Science and Technology · doi:10.1088/2058-9565/adc3ba · Cited by 11
Malhotra, A., & Gupta, S. K. (2025) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2025.3570193 · Cited by 3
“1) ADC Reference Optimization: In conventional CiM designs, the ADC quantization step size ( I quant ), which is the difference between successive ADC reference currents, is typically set to match the ON current of the bitcell ( I LRS ), assuming ideal behavior [45], [46].”
Efficient Reprogramming of Memristive Crossbars for DNNs: Weight Sorting and Bit Stucking
Farias, M., & Kung, H. T. (2025) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas56072.2025.11043370 · Cited by 2
Programmable soliton dynamics in all-Josephson-junction logic cells and networks
Ruzhickiy, V. I., Maksimovskaya, A. A., Bakurskiy, S. V., Schegolev, A. E., Tereshonok, M. V., Kupriyanov, M. Y., Klenov, N. V., & Soloviev, I. I. (2025) · Beilstein Journal of Nanotechnology · doi:10.3762/bjnano.16.131 · Cited by 3
Lü, S., Zhu, J., Zheng, L., Song, J., Hu, Z., Wang, X., & Wang, D. (2026) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2026.114074 · Cited by 4
Approximate Computing: Hardware and Software Techniques, Tools and Their Applications
Raza, M., Javed, S., Kazmi, M., Aziz, A., Haque, M. F. U., & Qazi, S. A. (2023) · Journal of Circuits Systems and Computers · doi:10.1142/s0218126624300010 · Cited by 2
Scalable Spintronic Synapses for Analog In-Memory Computing Based on Exchange-Coupled Nanostructures
Gupte, K. K., Mugdho, S. S., & Wang, C. (2024) · IEEE International Conference on Rebooting Computing (ICRC) · doi:10.1109/icrc64395.2024.10937014 · Cited by 2
“Crossbar in-memory computing (IMC) is an emerging computing architecture that is promising to achieve hardware acceleration of MVM computation [6].”
Design Tools for Resistive Crossbar based Machine Learning Accelerators
Chakraborty, I., Roy, S., Sridharan, S., Ali, M., Ankit, A., Jain, S., & Raghunathan, A. (2021) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas51828.2021.9458433 · Cited by 3
“To that effect, resistive crossbars based on Non-volatile memory (NVM) technologies promise to be a suitable candidate for building in-memory computing systems for ML workloads [8].”
BackgroundSelf-citationDube, A., Wagle, A., Singh, G., & Vrudhula, S. B. K. (2022) · doi:10.1145/3508352.3549385 · Cited by 2
“This has reinvigorated and accelerated the development of processing inmemory (PiM) architectures [10, 16, 32, 5, 6, 37, 4], which involves integrating processing elements inside the memory architecture so that the data can be processed in-place, avoiding the data transfers between the processor and memory.”
BackgroundMethodInfluentialSuper-steep synapses based on positive feedback devices for reliable binary neural networks
Kwon, D., Kim, H., Lee, K.-H., Hwang, J., Shin, W., Bae, J., Woo, S. Y., & Lee, J. (2023) · Applied Physics Letters · doi:10.1063/5.0131235 · Cited by 2
PyMem: A Graphical User Interface Tool for Neuromemristive Hardware–Software Co-Design
Radhakrishnan, A., Palliyalil, J., Babu, S., Dorzhigulov, A., & James, A. P. (2024) · IEEE Open Journal of the Industrial Electronics Society · doi:10.1109/ojies.2024.3363093 · Cited by 4
“Due to various device-and circuit-level variations, creating an ideal memristor-neural network is not a realistic solution for neuromorphic hardware applications [9], [10], [11].”
BackgroundApproximate Computing for Efficient Neural Network Computation: A Survey
Zhang, H., Asadikouhanjani, M., Han, J., Deivalakshmi, S., & Ko, S. (2012) · Springer eBooks · doi:10.1007/978-3-030-98347-5_16 · Cited by 2
Robustness Hidden in Plain Sight: Can Analog Computing Defend Against Adversarial Attacks?Preprint
Roy, D., Chakraborty, I., Ibrayev, T., & Roy, K. (2020) · arXiv (Cornell University) · Cited by 4
Self-citationReliable Analog In-Memory Computing with Crossbars: Memristors for Analog Neural Computing
James, A. P. (2025) · Foundations and Trends® in Integrated Circuits and Systems · doi:10.1561/3500000018 · Cited by 1
FairXbar: Improving the Fairness of Deep Neural Networks with Non-Ideal in-Memory Computing Hardware
Mugdho, S. S., Guo, Y., Rogers, E. G., Zhao, W., Shi, Y., & Wang, C. (2025) · Design, Automation, and Test in Europe · doi:10.23919/date64628.2025.10993038 · Cited by 2
“Crossbar-based IMC has emerged as a promising hardware platform for accelerating DNN processing [10], [13].”
Farias, M., & Kung, H. T. (2025) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas56072.2025.11043969 · Cited by 1
“By integrating storage and computation directly within the memory array, crossbars perform multiply-accumulate (MAC) operations with higher speed and lower power consumption compared to conventional digital systems [4]–[8].”
Stochastic nanomagnets as current digitizers for efficient probabilistic machine learning
Koh, D., Rogers, E. G., Gupte, K. K., Wang, Q., McGoldrick, B. C., Lee, J., Baldo, M. A., Wang, C., & Liu, L. (2025) · Physical Review Applied · doi:10.1103/s86j-5jcr · Cited by 2
PimShare: Scheduling for Multi-DNN Inference on Processing-in-Memory Accelerated Edge Server
Chen, X., Xie, Z.-L., Li, H., Chen, K., Shou, L.-D., Jiang, D., & Chen, G. (2026) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2025.3641876 · Cited by 1
“With the usage of nonvolatile memory (NVM) medium, such as resistive RAM (ReRAM) [66], spin-transfer torque RAM (STT-RAM) [5], or phase-change memory (PCM) [60], PIM accelerators could also overcome power leakage issues [58] and thus provide high storage and computation density [12], [53].”
Scalable and robust multi-bit spintronic synapses for analog in-memory computing
Gupte, K. K., Mugdho, S. S., Huang, C., & Wang, C. (2026) · npj Unconventional Computing · doi:10.1038/s44335-026-00055-7 · Cited by 4
“Crossbar in-memorycomputing(IMC)isanemergingcomputingarchitecturethatis promisingtoachievehardwareaccelerationofMVMcomputation 6 .”
RRAM-Based Precision-Scaleable Computing-In-Memory Scheme and Its Error Correction Approach
Ma, W., Li, L., Li, Z., Zhao, G., Long, X., & Huang, M. (2023) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas57966.2023.10168585 · Cited by 2
“However, it is inherently limited by the large device- and circuit-level non-idealities [10], causing large computation errors and high accuracy loss.”
BackgroundMemristive Devices for Neuromorphic and Deep Learning Applications
Walters, B., Lammie, C., Eshraghian, J. K., Yakopcic, C., Taha, T. M., Genov, R., Jacob, M. V., Amirsoleimani, A., & Azghadi, M. R. (2023) · Royal Society of Chemistry eBooks · doi:10.1039/bk9781839169946-00680 · Cited by 1
A Current-Mirroring Voltage Buffer for Analog Read-Out in Memristor Crossbar Arrays
Zoche, J., Heinen, R., Grobe, J., Wunderlich, R., & Heinen, S. J. (2024) · Conference on Ph.D. Research in Microelectronics and Electronics · doi:10.1109/prime61930.2024.10559676 · Cited by 1
“This allows even higher integration density and power efficiency than purely digital memristor crossbar arrays [4].”
BackgroundMethodModel quantization for computing-in-memory: a survey
Sun, S., Bai, J., Chen, H., Deng, K., Xie, Z., Li, J., Cao, B., Zhang, H., Kang, W., & Zhao, W. S. (2025) · Science China Information Sciences · doi:10.1007/s11432-024-4522-8 · Cited by 3
“Prior studies have shown that ADCs contribute over 60%–80% of the total energy consumption and chip area in CIM hardware [76,77].”
Sharda, J., Sharma, R., & Bhowmik, D. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401530 · Cited by 2
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.