Mask Technique for Fast and Efficient Training of Binary Resistive Crossbar Arrays
The DOI matches its Crossref record; 29 of the 30 references it cites were found in a registry or an index.
Abstract
Resistive crossbar arrays, despite their advantages such as parallel processing, low power, and high density, suffer from the sneak path problem created by IR voltage drops. The crossbar non-idealities considerably degrade the accuracy of matrix multiplication operation, which is the cornerstone of hardware accelerated neural networks. In this paper, we propose a novel technique to capture the sneak path problem to enable off-chip learning and weight transfer. The proposed technique can be easily integrated into any neural network training framework. Performance results show a significant improvement after retraining the network with the proposed mask technique. Two mask solutions have been proposed and studied to capture the sneak path problem in resistive crossbar arrays. Both mask solutions were successfully able to achieve classification accuracy close to the baseline accuracy. The tradeoffs between the two solutions are discussed and compared in terms of accuracy, power, and area.
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
“This representation method, shown in Fig. 6d, has been chosen in this study because it doubles the range of conductance levels of the crossbar, making it less susceptible to noise and variability104.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [104]
In the citing paper
“A widely accepted104,271, alternative design to minimize this problem consists in dividing the large crossbars into smaller ones (Supplementary Fig. 9b), whose reduced size improves their read margin (that is the portion of the applied voltage in the inputs that is actually delivered to the memristors).”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [104]
In the citing paper
“By exploding the integrability of the crossbar with CMOS circuitry, vertical interconnects used to connect the outputs of the vertical crossbar partitions may be placed under the partitioned structure (as well as the analogue sensing electronics) allowing the partitioned crossbar to maintain a similar area consumption than the original non-partitioned case104.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [104]
Citations
Cites
Training and operation of an integrated neuromorphic network based on metal-oxide memristors
Prezioso, M., Merrikh-Bayat, F., Hoskins, B. D., Adam, G. C., Likharev, K. K., & Strukov, D. B. (2015) · Nature · doi:10.1038/nature14441 · Cited by 3,001 · cited 1×
“In order to create negative weights, two weight realization techniques have been proposed: (i) using two RRAM cells per weight [11] as shown in Fig.”
MethodHubara, I., Soudry, D., & Yaniv, R. E. (2016) · arXiv · doi:10.48550/arxiv.1602.02505 · Cited by 1,667 · cited 5×
“To evaluate the accuracy of our RRAM crossbar-based DNN hardware, we use the BinaryNet framework [10] extended with our RRAM crossbar array simulation module.”
BackgroundMethodInfluentialISAAC: 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 1×
“eliminating the need for analog to digital and digital to analog converters or sophisticated analog circuits [8], in addition to allowing bitwise communication between the layers.”
BackgroundEyeriss: a spatial architecture for energy-efficient dataflow for convolutional neural networks
Chen, Y.-H., Emer, J., & Sze, V. (2016) · ACM SIGARCH Computer Architecture News · doi:10.1145/3007787.3001177 · Cited by 1,182
Neuro-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×
“While the previous work in this area typically uses a simplified model of crossbar arrays, in order to tackle large applications such as deep neural networks (DNNs), it is important to consider physical realities or RRAM’s non-idealities such as nonlinear and asymmetric potentiation and depression, device variability,…”
BackgroundEfficient and self-adaptive in-situ learning in multilayer memristor neural networks
Li, C., Belkin, D., Li, Y., Yan, P., Hu, M., Ge, N., Jiang, H., Montgomery, E., Lin, P., Wang, Z., et al. (2018) · Nature Communications · doi:10.1038/s41467-018-04484-2 · Cited by 931
Memristor‐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 1×
“e device’s resistance is voltage independent) [28].”
Recommended Methods to Study Resistive Switching Devices
Lanza, M., Wong, H. ‐. P., Pop, E., Ielmini, D., Strukov, D., Regan, B. C., Larcher, L., Villena, M. A., Yang, J. J., Goux, L., et al. (2019) · Advanced Electronic Materials · doi:10.1002/aelm.201800143 · Cited by 674 · cited 1×
“Furthermore, in terms of devices, bipolar RRAMs have shown promising performance in terms of endurance, variability and energy [9].”
BackgroundBayat, F. M., Prezioso, M., Chakrabarti, B., Nili, H., Kataeva, I., & Strukov, D. (2018) · Nature Communications · doi:10.1038/s41467-018-04482-4 · Cited by 377
Size-Dependent Resistivity in Nanoscale Interconnects
Josell, D., Brongersma, S. H., & Tőkei, Z. (2009) · Annual Review of Materials Research · doi:10.1146/annurev-matsci-082908-145415 · Cited by 286 · cited 1×
“Thus, the resistivity is wire size dependent especially with wire width less than 100 nm [25].”
Background
Research works printed but not linked: 1
- wilson, 2013, International technology roadmap for semiconductors (itrs), Semiconductor Industry Association
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
“2. where g ref stands for the 0.5· G LRS conductances of the reference column and g 0 i,j is calculated in such a way that devices with a conductance above 0.5· G LRS produce positive synaptic weights, and those with a conductance below 0.5· G LRS produce negative synaptic weights 104 .”
InfluentialSelf-citationLearning to Predict IR Drop with Effective Training for ReRAM-based Neural Network Hardware
Lee, S., Jung, G., Fouda, M. E., Lee, J., Eltawil, A. M., & Kurdahi, F. J. (2020) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18072.2020.9218735 · Cited by 51
“Mask [11] assumes (2) and that h2 is an element-wise multiplication.”
BackgroundMethodResultInfluentialSelf-citationIR-QNN Framework: An IR Drop-Aware Offline Training of Quantized Crossbar Arrays
Fouda, M. E., Lee, S., Lee, J., Kim, G. H., Kurdahi, F. J., & Eltawil, A. M. (2020) · IEEE Access · doi:10.1109/access.2020.3044652 · Cited by 44
“Larger array sizes lead to worse IR drops, causing higher degradation in performance, as discussed in [26].”
BackgroundMethodInfluentialSelf-citationAguirre, F. L., Pazos, S. M., Palumbo, F., Sune, J., & Miranda, E. (2020) · IEEE Access · doi:10.1109/access.2020.3035638 · Cited by 36
“…vertical interconnects used to connect the outputs of the vertical CPA partitions may be placed under the partitioned struc-FIGURE ture, as well as the analogue sensing electronics, allowing the partitioned CPA to maintain a similar area consumption than the original non-partitioned case [73].”
MethodOn-Chip Error-Triggered Learning of Multi-Layer Memristive Spiking Neural Networks
Payvand, M., Fouda, M. E., Kurdahi, F. J., Eltawil, A. M., & Neftci, E. (2020) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2020.3040248 · Cited by 38
“Partitioning the larger layers into smaller memory banks and summing up the output current can be employed as a solution for scaling up RCAs [72], [73].”
MethodSelf-citation
Metadata
- Authors
- Fouda, M. E., Lee, S., Lee, J., Eltawil, A., & Kurdahi, F.
- Affiliations
- University of California, Irvine (US)
- Ulsan National Institute of Science and Technology (KR)
- Venue
- IEEE Transactions on Nanotechnology
- Published
- 2019
- Volume
- 18
- Pages
- 704-716
- ISSN
- 1536-125X (ISSN-L)
- 1941-0085
- OpenAlex
- W2957706073
- Cited by
- 41
- References
- 30
- Citations by year
- 2026: 1 · 2025: 5 · 2024: 5 · 2023: 4 · 2022: 9 · 2021: 6 · 2020: 11
- FWCI
- 2.34
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Transactions on Nanotechnology (publishedVersion)
- Scholarworks@UNIST (Ulsan National Institute of Science and Technology) (submittedVersion)
- Funders
- National Science Foundation (1710746)
- National Research Foundation of Korea
- Ministry of Science, ICT and Future Planning (2016M3A7B4909668)
- Basic Science Research Program through NRF
- Ministry of Education (NRF-2017R1D1A1B03033591)
- Free Innovative Research Fund (1.170067.01)
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Memory and Neural Computing
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Low-power high-performance VLSI design
- Keywords
- sneak paths, weight transfer, matrix multiplication, in-memory computing, neural network retraining
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Mask Technique for Fast and Efficient Training of Binary Resistive Crossbar Arrays. https://app.cochat.ai/papers/10.1109/tnano.2019.2927493
- The paper
- Fouda, M. E., Lee, S., Lee, J., Eltawil, A., & Kurdahi, F. (2019). Mask Technique for Fast and Efficient Training of Binary Resistive Crossbar Arrays. IEEE Transactions on Nanotechnology. https://doi.org/10.1109/tnano.2019.2927493
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.