On the Accuracy of Analog Neural Network Inference Accelerators
Open accessThe DOI matches its Crossref record; 68 of the 71 references it cites were found in a registry or an index.
Abstract
Specialized accelerators have recently garnered attention as a method to reduce the power consumption of neural network inference. A promising category of accelerators utilizes nonvolatile memory arrays to both store weights and perform in situ analog computation inside the array. While prior work has explored the design space of analog accelerators to optimize performance and energy efficiency, there is seldom a rigorous evaluation of the accuracy of these accelerators. This work shows how architectural design decisions, particularly in mapping neural network parameters to analog memory cells, influence inference accuracy. When evaluated using ResNet50 on ImageNet, the resilience of the system to analog non-idealities—cell programming errors, analog-to-digital converter resolution, and array parasitic resistances—all improve when analog quantities in the hardware are made proportional to the numerical values in the network. Moreover, contrary to the assumptions of prior work, nearly equivalent resilience to cell imprecision can be achieved by fully storing weights as analog quantities, rather than spreading weight bits across multiple devices, often referred to as bit slicing. By exploiting proportionality, analog system designers have the freedom to match the precision of the hardware to the needs of the algorithm, rather than attempting to guarantee the same level of precision in the intermediate results as an equivalent digital accelerator. This ultimately results in an analog accelerator that is more accurate, more robust to analog errors, and more energy-efficient.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
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?
The cited source doesn’t support the claim as written.
In the citing paper
“ADCs with at least 8-bit resolution are necessary to achieve high (>90%) classification accuracy in a ResNET50-1.5 ANN used to classify the ImageNET157 database”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [157]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 1×
“The sensitivity to cell errors (described more fully in Section 5) differs dramatically for networks trained on three datasets—ImageNet [15], CIFAR10 [36], and MNIST [37]—with ImageNet being by far the most sensitive.”
BackgroundRethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138
Learning Multiple Layers of Features from Tiny Images
Krizhevsky, A. (2024) · doi:10.57702/zp44cu3g · Cited by 23,395
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017) · arXiv · doi:10.48550/arxiv.1704.04861 · Cited by 19,015
Han, S., Mao, H., & Dally, W. J. (2015) · arXiv · doi:10.48550/arxiv.1510.00149 · Cited by 4,662
In-Datacenter Performance Analysis of a Tensor Processing Unit
Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., et al. (2017) · Proceedings of the 44th Annual International Symposium on Computer Architecture · doi:10.1145/3079856.3080246 · Cited by 4,537
Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference
Jacob, B., Kligys, S., Chen, B., Zhu, M., Tang, M., Howard, A., Adam, H., & Kalenichenko, D. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00286 · Cited by 4,114
End-to-end arguments in system design
Saltzer, J. H., Reed, D. P., & Clark, D. D. (1984) · ACM Transactions on Computer Systems · doi:10.1145/357401.357402 · Cited by 2,538 · cited 1×
“[53], this robustness allows hardware requirements (such as the array size and the analog-to-digital converter resolution) to be relaxed while still ensuring high end-to-end accuracy of an application neural network.”
Background
Research works printed but not linked: 8
- chollet, 2015, Keras
- murmann, 2020, ADC Performance Survey 1997–2020
- agarwal, 2017, CrossSim
- 2019, MLPerf Inference V0 5 NVIDIA-Optimized Implementations for Open Division
- 2019, MLPerf Inference ImageNet Calibration Set
- 2019, MLPerf Inference Benchmark Suite
- lecun, 2021, The MNIST Database of Handwritten Digits
- hayakawa, 2015, Highly reliable TaOx ReRAM with centralized filament for 28-nm embedded application, Proc Symp VLSI Technol, 14t
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
Andrulis, T., Emer, J., & Sze, V. (2023) · Proceedings - International Symposium on Computer Architecture/Proceedings · doi:10.1145/3579371.3589062 · Cited by 70
“Offset encoding provides noise resistance [73], and Center+Offset increases bit sparsity and decreases noise.”
BackgroundMethodInfluentialXiao, T. P., Feinberg, B., Bennett, C. H., Agrawal, V. S., Saxena, P., Prabhakar, V., Ramkumar, K., Medu, H., Raghavan, V., Chettuvetty, R., et al. (2022) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2021.3134313 · Cited by 46
Self-citationCiMLoop: A Flexible, Accurate, and Fast Compute-In-Memory Modeling Tool
Andrulis, T., Emer, J., & Sze, V. (2024) · IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS) · doi:10.1109/ispass61541.2024.00012 · Cited by 60
“2) Representation: How data values appear is determined by how the hardware represents operands [47].”
BackgroundAnalog Multiply-Accumulate Cell With Multi-Bit Resolution for All-Analog AI Inference Accelerators
Nägele, R., Finkbeiner, J., Stadtlander, V., Grözing, M., & Berroth, M. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2023.3268728 · Cited by 27
“property of neural networks supports this statement [9].”
BackgroundInfluentialQuantized 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
Progress and challenges of phase change memory for in-memory computing
Su, X., Guo, Z., Zhang, G., Tsai, H.-Y. S., & Li, N. (2025) · Current Opinion in Solid State and Materials Science · doi:10.1016/j.cossms.2025.101225 · Cited by 14
Towards Green AI: Current Status and Future Research
Clemm, C., Stobbe, L., Wimalawarne, K., & Druschke, J. (2024) · Electronics Goes Green · doi:10.23919/egg62010.2024.10631247 · Cited by 39
“Specialized accelerators for AI have the potential to further reduce the power consumption of neural networks, including analog accelerators, a category of devices utilizing non-volatile memory to store weights and perform in-situ analog computation [69].”
BackgroundGraphene-Based Artificial Dendrites for Bio-Inspired Learning in Spiking Neuromorphic Systems
Liu, S., Akinwande, D., Kireev, D., & Incorvia, J. A. C. (2024) · Nano Letters · doi:10.1021/acs.nanolett.4c00739 · Cited by 15
Spear, M. E., Kim, J. E., Bennett, C. H., Agarwal, S., Marinella, M. J., & Xiao, T. P. (2023) · IEEE Journal on Exploratory Solid-State Computational Devices and Circuits · doi:10.1109/jxcdc.2023.3315134 · Cited by 12
“To date, the ADC’s effect on neural network accuracy has only been analyzed through the lens of generic models for ADC quantization, clipping, and output noise [10], [11], [12], [13], [14], [15], [16], [17].”
BackgroundSelf-citationTowards Reliable and Energy-Efficient RRAM Based Discrete Fourier Transform Accelerator
Wen, J., Baroni, A., Pérez, E., Uhlmann, M., Fritscher, M., KrishneGowda, K., Ulbricht, M., Wenger, C., & Krstic, M. D. (2024) · Design, Automation, and Test in Europe · doi:10.23919/date58400.2024.10546709 · Cited by 10
“Mapping the signed matrices to the crossbar with the offset approach may eventually degrade the robustness of the system against the device variabilities and, as a consequence, deteriorate the computational accuracy [9], [10].”
BackgroundMethodKho, W., Hwang, H., Noh, T., Kim, H., Lee, J. M., & Ahn, S. (2024) · Advanced Intelligent Systems · doi:10.1002/aisy.202400211 · Cited by 8
Vignali, R., Zurla, R., Pasotti, M., Rolandi, P. L., Singh, A., Gallo, M. L., Sebastian, A., Jang, T., Antolini, A., Scarselli, E. F., et al. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3340112 · Cited by 13
“A concept introduced in [20] highlights the fact that the computation of the inference of DNNs in AiMC architecture is more accurate when the conductance of each PCM cell is directly proportional to the weights of the kernels of the NN and the analog result of the MAC operation is directly…”
BackgroundResultNeuromorphic Hebbian learning with magnetic tunnel junction synapses
Zhou, P., Edwards, A. J., Mancoff, F. B., Aggarwal, S., Heinrich-Barna, S. K., & Friedman, J. S. (2025) · Communications Engineering · doi:10.1038/s44172-025-00479-2 · Cited by 18
“The analog and multilevel states of memristor and PCM devices exhibit stochastic writing, weight drift over time, and limited endurance that reduces the precision of synapse weights, leading to drastic decreases in system accuracy as thoroughly shown in [37].”
Dual-ion ECRAM as a stable and accurate analog synapse
Faruque, H. M. R., Islam, M. A., Voegtle, M. J., Holtzman, N., Bae, I.-T., Swain, S. S., Beom, K., Dempsey, R., Smith, C. J., Xiao, T. P., et al. (2025) · Device · doi:10.1016/j.device.2025.100928 · Cited by 9
Self-citationXiao, S., Li, F., Hao, T., Xiao, L., Xiao, M., Mao, W., & Han, G. (2025) · Integrated Circuits and Systems · doi:10.23919/ics.2025.3571019 · Cited by 7
Bringing In-Sensor Intelligence in Radiation Detectors: a Short Review
Carminati, M., Giacomo, S. D., Ronchi, M., Borghi, G., & Fiorini, C. E. (2023) · IEEE EUROCON - International Conference on Computer as a Tool (EUROCON) · doi:10.1109/eurocon56442.2023.10198991 · Cited by 6
“Analog accelerators [6] have been emerging as promising alternatives to digital engines, for instance adopting different sensing paradigms such as stochastic resonance [7] or emerging technologies like cross-bar arrays of memristors and non-volatile memories for in-memory computing (IMC) [8], [9].”
BackgroundA Calibratable Model for Fast Energy Estimation of MVM Operations on RRAM Crossbars
Cubero-Cascante, J., Vaidyanathan, A. M., Pelke, R., Pfeifer, L., Leupers, R., & Joseph, J. M. (2024) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas59952.2024.10595858 · Cited by 6
“The bias G b is set to ( G min + G max ) / 2 [ differential mapping , the weight values are represented with a pair of values, so that Typically, G ij is set to G min for positive numbers and G ij is set to G min for negative numbers [ 26].”
BackgroundTotal Ionizing Dose Response of 128 Analog States in Computational Charge-Trap Memory
Xiao, T. P., Wilson, D., Bennett, C. H., Hughart, D. R., Feinberg, B., Agrawal, V. S., Puchner, H., Marinella, M. J., & Agarwal, S. (2023) · IEEE Transactions on Nuclear Science · doi:10.1109/tns.2023.3333812 · Cited by 9
“As with other types of analog errors [27], the complexity of the application has a strong effect on the severity of accuracy loss caused by radiation-induced perturbations to the weight values.”
BackgroundMethodInfluentialSelf-citationИгоревич, А. В., Tominov, R. V., Vakulov, Z., Rodriguez, D. J., Polupanov, N. V., & Smirnov, V. A. (2025) · Nanomaterials · doi:10.3390/nano15010075 · Cited by 5
“One of the solutions to this problem is the use of additional accelerators that deal with neural network calculations [14–18].”
Multistate resistance in TaN/(Hf,Zr)O2/Ta ferroelectric tunnel junctions
Henry, M. D., Stoltzfus, D., Lenox, M. K., Jaszewski, S. T., Young, T. R., & Ihlefeld, J. F. (2025) · Journal of Applied Physics · doi:10.1063/5.0256686 · Cited by 5
Enabling High-Speed, High-Resolution Space-based Focal Plane Arrays with Analog In-Memory Computing
Xiao, T. P., Wahby, W. S., Bennett, C. H., Hays, P., Agrawal, V., Marinella, M. J., & Agarwal, S. (2023) · IEEE Symposium on VLSI Technology and Circuits (VLSI Technology and Circuits) · doi:10.23919/vlsitechnologyandcir57934.2023.10185348 · Cited by 7
Self-citationSurvey of Security Issues in Memristor-Based Machine Learning Accelerators for RF Analysis
Lillis, W., Hoffing, M. C., & Burleson, W. P. (2024) · Chips · doi:10.3390/chips3020009 · Cited by 5
Overcoming the noise in neural computing
Aimone, J. B., & Agarwal, S. (2024) · Science · doi:10.1126/science.adn8545 · Cited by 7
Self-citationWilson, D., Dempsey, R., Xiao, T. P., Spear, M. E., Hughart, D. R., Bennett, C. H., Agrawal, V. S., Puchner, H., Agarwal, S., & Marinella, M. J. (2025) · IEEE Transactions on Nuclear Science · doi:10.1109/tns.2025.3543349 · Cited by 5
“In SONOS, it has been found to be advantageous from the perspective of accuracy and energy efficiency to map weights to subthreshold currents [11].”
Self-citationX-Ray Irradiation Improved WSe2 Optical–Electrical Synapse for Handwritten Digit Recognition
Chen, C., Sun, Q., Lu, Y., & Chen, P. (2025) · Nanomaterials · doi:10.3390/nano15181408 · Cited by 4
“To validate the device’s applicability in neuromorphic computing, as shown in Figure 5a, we simulated a neural network using the CrossSim crossbar simulator with the MNIST dataset [61,62].”
Bommana, A. R., Bishnoi, R., Karimi, N., Firouzi, F., & Chakrabarty, K. (2026) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2026.3657612 · Cited by 5
“Variation-aware and drift-aware retraining schemes have been proposed to adapt model weights to device-level non-idealities and gradual conductance drift [63], [64].”
Bommana, A. R., Feinberg, B., Xiao, P., Bennet, C., Marinella, M. J., & Chakrabarty, K. (2026) · Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1109/asp-dac66049.2026.11420230 · Cited by 3
“However, analog computations are inherently susceptible to various sources of noise, such as device-to-device variation, cycle-to-cycle variation, conductance drift over time, and environmental factors such as temperature or radiation-induced drift [5], [7]–[9].”
InfluentialSelf-citationCharacterization of an Analog MAC Cell with Multi-Bit Resolution for AI Inference Accelerators
Nägele, R., Finkbeiner, J., Grözing, M., Berroth, M., & Rademacher, G. (2024) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas59952.2024.10595889 · Cited by 3
The Effect of Heavy-Ion Strikes on Charge Trap Memory Arrays With Analog State Programmability
Xiao, T. P., Bennett, C. H., Dempsey, R., Wilson, D., Joffrion, J. B., Udoni, D. M., Talin, A. A., Agrawal, V. S., Puchner, H., Marinella, M. J., et al. (2024) · IEEE Transactions on Nuclear Science · doi:10.1109/tns.2024.3505962 · Cited by 3
“Unlike digital data storage, analog IMC can benefit from the error tolerance of the application algorithm and does not require its analog state distributions to be strongly separable [9].”
Self-citationBi, J.-H., Yang, S., Zhang, P., & Chen, S. (2024) · Globecom · doi:10.1109/globecom52923.2024.10901048 · Cited by 5
“The conductance errors can be modeled as Gaussian random variables with mean 0 and variance σ 2 m [13].”
Bi, J.-H., Yang, S., Zhang, P., & Chen, S. (2024) · Globecom · doi:10.1109/globecom52923.2024.10901825 · Cited by 3
“Therefore, let v 1 be the output voltages of the set of OAs, we have: where i − denotes the currents flowing into the inverting-input nodes of OAs.”
Harmonica: Hybrid Accelerator to Overcome Imperfections of Mixed-signal DNN Accelerators
Behnam, P., Kamal, U., Shafiee, A. R., Tumanov, A., & Mukhopadhyay, S. (2024) · IEEE International Parallel and Distributed Processing Symposium (IPDPS) · doi:10.1109/ipdps57955.2024.00061 · Cited by 6
“Xiao et al. explore the source of errors in analog accelerators and demonstrate how existing architectural choices impact accuracy [47].”
BackgroundIn Situ Analog In-Memory Computing Under Ionizing Radiation Exposure
Xiao, T. P., Siath, M., Spear, M. E., Wilson, D., Bennett, C. H., Feinberg, B., Hughart, D. R., Neuendank, J., Brown, W. E., Barnaby, H., et al. (2025) · IEEE Transactions on Nuclear Science · doi:10.1109/tns.2025.3537985 · Cited by 2
“Some prior work on analog IMC has found that inference accuracy can be a strong function of the weight distribution shape, depending on the specific characteristics of the dominant analog errors [31], [34].”
Self-citationConstraints on Bit Precision and Row Parallelism for Reliable Computing-in-Memory
Li, Y., Wang, S., & Sun, Z. (2025) · Electronics · doi:10.3390/electronics14132532 · Cited by 4
ReSpike: A Co-Design Framework for Evaluating SNNs on ReRAM-Based Neuromorphic Processors
Asifuzzaman, K., Young, A. R., Date, P., Kulkarni, S., Miniskar, N. R., Marinella, M. J., & Vetter, J. S. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-99857-7_13 · Cited by 3
Self-citationXu, J., Zheng, Y., Shen, R., Wang, R., Li, J., Wang, D., Stathis, D., Zheng, L., Zou, Z., & Hemani, A. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.132107 · Cited by 4
Lin, Z., Wu, X., Chen, Z., Zhang, J., Deng, J., & Feng, Y. (2026) · iScience · doi:10.1016/j.isci.2026.114642 · Cited by 4
MoF-LoRA: Mixture of Low-Rank Fault-Tolerant Experts for RRAM-Based In-Memory Computing
Ahmed, S. T., Ortega, E., Xiao, P., Feinberg, B., Bennett, C. H., Marinella, M. J., & Chakrabarty, K. (2026) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2026.3655242 · Cited by 3
“1) Programming Error: Programming errors in RRAM refer to inaccuracies in setting target conductance values, arising primarily from device-level non-idealities such as process variations [17].”
InfluentialSelf-citationA CMOS Implementation of a Switched Capacitor Analog Neural Network ASIC (ANNA)
Giacomo, S. D., Ronchi, M., Borghi, G., Carminati, M., & Fiorini, C. E. (2023) · IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC) · doi:10.1109/nssmicrtsd49126.2023.10338747 · Cited by 1
Effect of Parameter Drift on the Accuracy of Analog CNN Accelerators
Keyser, M., Ordubadi, F., & Tajalli, A. (2024) · IEEE International Midwest Symposium on Circuits and Systems (MWSCAS) · doi:10.1109/mwscas60917.2024.10658938 · Cited by 1
“These variations in the physical and electrical properties of components are categorized as global and local variations [8]–[11].”
BackgroundImproving Processing-In-Memory Chip Inferencing Accuracy Through Noise-Aware Training
Kim, Y., Lu, C., Qiu, X., & Zhao, Y. (2025) · World Conference on Computing and Communication Technologies (WCCCT) · doi:10.1109/wccct65447.2025.11027943 · Cited by 1
“Numerous efforts have been made to mitigate the accuracy loss in PIM chips [9-12].”
Multi-Level Analog Computing-In-Memory FeFET-based Unit Cell for Deep Learning
Pereira-Rial, O., Dahlberg, H., García-Lesta, D., Brea, V. M., López, P., Cabello, D., & Wernersson, L. (2025) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas56072.2025.11043743 · Cited by 1
“Also, the fact that the unit cell features a differential mode preserves weight proportionality, leading to higher accuracy levels [12].”
Wang, Q., Tamukoh, H., & Morie, T. (2025) · Frontiers in Neuroscience · doi:10.3389/fnins.2025.1656892 · Cited by 3
Fault Tolerance in RRAM-based AI Accelerator with Guided Randomized Activation
Ahmed, S. T., Ortega, E., Depsey, R., Xiao, P., Feinberg, B., Bennett, C. H., Marinella, M. J., & Chakrabarty, K. (2025) · IEEE International Test Conference (ITC) · doi:10.1109/itc58126.2025.00039 · Cited by 2
“Existing solutions attempt to mitigate these effects through error correction mechanisms [13], noise-aware offline training [10], [14], in-situ training [15], and per-device re-calibration for variation-tolerance [14].”
InfluentialSelf-citationSerm, T.-B., Chen, T.-C., Wang, P.-K., Wu, Y., Chan, T.-Y., Chiang, M., & Ho, Y.-H. (2025) · Applied Physics Letters · doi:10.1063/5.0304098 · Cited by 1
Li, Y., Zheng, X., Zhang, W., Wang, H., Chen, T., Cui, X., He, X., Ly, T. H., & Zhao, J. X. (2025) · Small · doi:10.1002/smll.202503908 · Cited by 2
HAT-FI: Hardware-Aware Training for Fault-Tolerant LLM Inference on RRAM-Based Compute-in-Memory
Sharma, P., Ahmed, S. T., Firouzi, F., & Chakrabarty, K. (2026) · IEEE VLSI Test Symposium (VTS) · doi:10.1109/vts69484.2026.11563304 · Cited by 1
“…degrade inference accuracy throughout the device’s operational lifetime: (1) programming errors from stochastic filament formation at write time [8], (2) conductance drift from post-programming relaxation [9], (3) stuck-at faults from endurance degradation [10], (4) stochastic read noise per…”
Physics-Driven In-House Memristor Crossbar Array Model for Image Classification
Canle, D. V., Mohammadidoghozloo, M., Costa, J. D., Leborán, V., Pardo, F., Pereira-Rial, O., Rivadulla, F., Sánchez, V. M. B., & López, P. (2026) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas66217.2026.11562579 · Cited by 1
Boosting the Memory Window of Memristive Stacks via Engineered Interfaces with High Ionic Mobility
Varela‐Domínguez, N., Canle, D. V., Davey-García, N., Leborán, V., Ramos, R., Casanova, F., Hueso, L. E., Brea, V. M., Lopez, P., Rivadulla, F., et al. (2026) · ACS Applied Electronic Materials · doi:10.1021/acsaelm.6c01143 · Cited by 1
“To be checked”
ReFFT: An Energy-Efficient RRAM-Based FFT Accelerator
Wen, J., Baroni, A., Uhlmann, M., Pérez, E., Wenger, C., & Krstic, M. D. (2025) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2025.3627146 · Cited by 1
“An alternative offset-based method shifts all coefficients into the positive region, but it often reduces computational accuracy [21], [22].”
InfluentialSEC-CiM+: Selective and Reconfigurable Error Compensation for ReRAM-based Compute-in-Memory
Bommana, A. R., Firouzi, F., Ogbogu, C., Joardar, B. K., Doppa, J. R., Pande, P. P., & Chakrabarty, K. (2026) · ACM Transactions on Design Automation of Electronic Systems · doi:10.1145/3812655 · Cited by 1
Analog Neural Network Inference Accuracy in One-Selector One-Resistor Memory Arrays
Kim, J. E., Xiao, T. P., Bennett, C. H., Wilson, D., Spear, M. E., Siath, M., Feinberg, B., Agarwal, S., & Marinella, M. J. (2022) · IEEE International Conference on Rebooting Computing (ICRC) · doi:10.1109/icrc57508.2022.00012
Self-citationModelling and Performance Assessment of a CMOS Switched Capacitor Analog Neuron
Ronchi, M., Giacomo, S. D., Amadori, M., Borghi, G., Carminati, M., & Fiorini, C. (2024) · IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC) · doi:10.1109/nss/mic/rtsd57108.2024.10655303
NavCim: Comprehensive Design Space Exploration for Analog Computing-in-Memory Architectures
Park, J., Kim, B. J., & Sung, H. (2024) · International Conference on Parallel Architectures and Compilation Techniques (PACT) · doi:10.1145/3656019.3676946
Architecture-Compiler Co-design for ReRAM-Based Multi-core CIM Architectures
Pelke, R., Bosbach, N., Degener, N., Cubero-Cascante, J., Staudigl, F., Leupers, R., & Joseph, J. M. (2024) · IFIP advances in information and communication technology · doi:10.1007/978-3-031-70947-0_2
Analog In-Memory Computing for the Synthetic Aperture Radar Polar Format Algorithm
Richardson, D. K., Xiao, T. P., West, R. D., Bennett, C. H., & Agarwal, S. (2025) · IEEE Transactions on Radar Systems · doi:10.1109/trs.2025.3570977
“The standard deviations, ∆ G , for the state-independent and state-dependent errors are described by [31] ∆ G ind = α ind G max (25) ∆ G dep = α dep G (26) where α is a user-specified parameter that quantifies the magnitude of each type of conductance error, G is the target (ideal) conductance of a…”
Self-citationMemory Arrays for FMCW Radar Signal Processing:A Feasibility Analysis
Anzum, M. S., & Menon, A. (2025) · The ... IEEE Annual Wireless and Microwave Technology Conference · doi:10.1109/wamicon64429.2025.11004117
“The current is then transformed into output voltages via summing amplifiers or transimpedance amplifiers and digitized using analog-to-digital converters (ADCs) [11].”
InfluentialRadiation Effects Modeling and Analysis in RRAM-Based In-Memory Computing Systems
Ren, T., Han, X., Yao, K., Guo, H., Wang, L., Zhao, Y., & Bi, W. (2025) · International Conference on Radiation Effects of Electronic Devices (ICREED) · doi:10.1109/icreed65908.2025.11036223
“…from each array can be represented by the output current: Recent studies have introduced architecture simulator platforms that support system-level design of IMC accelerators [4-6], accounting for device-level non-idealities such as resistance variations, on/off ratio, and conductance drift.”
Ren, T., Bi, W., Wang, L., & Zhao, Y. (2025) · 2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD) · doi:10.1109/iccad66269.2025.11240930
“Empirically, clipping these outliers has a negligible impact on model accuracy [9].”
InfluentialLiu, J., Tong, W., Wu, B., Cheng, H., Zhou, H., Wei, X., & Feng, D. (2025) · ACM Transactions on Architecture and Code Optimization · doi:10.1145/3777381
Reconfigurable Analog Neural Networks: Architecture, Design, and Performance Evaluation
Elhossan, A., Parent, S., Suthar, S., Turco, A., Zartash, H., & Elamien, M. B. (2025) · IEEE International Midwest Symposium on Circuits and Systems (MWSCAS) · doi:10.1109/mwscas53549.2025.11244495
Sakemi, Y., Awano, H., & Morie, T. (2025) · IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences · doi:10.1587/transfun.2025gci0001 · Cited by 2
Guo, L., Zhu, Z., Lin, Q., Xie, Y., Yang, H., Fu, W., & Wang, Y. (2025) · doi:10.1109/iccad66269.2025.11240851
“…above simulation paradigm, the absolute accuracy of the neural network on analog PIM architectures can be accurately evaluated, facilitating the analysis of various errors’ impacts on accuracy [17] and further optimization of algorithmic performance on the analog PIM architectures [9], [18]–[22].”
RRAM-Based Spectral-Domain Convolution Accelerator for Reliable and Energy-Efficient CNN Inference
Wen, J., Baroni, A., Wenger, C., Krstic, M. D., & Poehls, L. M. B. (2025) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2025.3644141
“This problem becomes more pronounced in crossbar-based VMMs, where accumulated outputs amplify the e ff ect of variability [21].”
A Blueprint for Accurate, Energy-Efficient DNN Inference via Capacitive In-Memory Processing
Demasius, K.-U., Lowa, A., & Murmann, B. (2025) · 2025 Cross-Disciplinary Conference on Memory-Centric Computing (CCMCC) · doi:10.1109/ccmcc67628.2025.11380554
“For charge trap memories, like SONOS transistors, a variation of σ ID = 20 nA for a drain current of I D = 400 nA were obtained [82].”
Let Analog Be Analog: Principles for Designing Analog Accelerators
Feinberg, B., Xiao, T. P., Bennett, C. H., & Agarwal, S. (2026) · IEEE Micro · doi:10.1109/mm.2026.3666132
Self-citationLearnable Center-Based Quantization for Efficient Analog PIM with Reduced ADC Precision
Yeon, S., & Ko, J. H. (2026) · Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1109/asp-dac66049.2026.11420293
“Additionally, differential encoding [19] is applied to align numerical values with analog hardware properties, enabling direct accuracy comparisons across different ADC precision levels.”
Predictive first-principles simulations for co-designing next-generation energy-efficient AI systems
Mamaluy, D., Udoy, M. R. I., Mendez, J. P., Feinberg, B., Pan, W., & Aziz, A. (2026) · Applied Physics Letters · doi:10.1063/5.0333458
Self-citationMixed-Precision Training and Compilation for RRAM-based Computing-in-Memory Accelerators
Pelke, R., Klein, J., Cubero-Cascante, J., Bosbach, N., Joseph, J. M., & Leupers, R. (2026) · Design, Automation, and Test in Europe · doi:10.23919/date69613.2026.11539262 · Cited by 1
“To map such 8 bit models onto low-resolution crossbars, two techniques are used: weight bit slicing and input bit slicing [21].”
Tian, Y., Sheheryar, T., Wang, S., & Lv, B. (2026) · Measurement · doi:10.1016/j.measurement.2026.122604
Demonstration of a Real-Time Floating-Point Photonic Accelerator and its Applications
Islam, M. S., Fardoost, A., Vanani, F. G., Desai, D., Sunder, S., Sobhanan, A., & Li, G. (2026) · doi:10.1364/cleo_si.2026.stu1e.6
CIM-FI: A HW-Aware Fault Injection Framework for Digital Compute-In-Memory DNN Accelerators
Chaidos, P., Maras, A., Spyrou, T., Gebregiorgis, A., Hamdioui, S., Soudris, D., & Xydis, S. (2026) · IEEE International On-Line Testing Symposium (IOLTS) · doi:10.1109/iolts69666.2026.11633664
“A large number of reliability studies in CIM are focused on ACIM [2], [3], [20]–[23].”
Analogue fast Fourier transforms for scalable and efficient signal processing
Xiao, T. P., Feinberg, B., Richardson, D. K., Cannon, M. J., Madsen, C., Medu, H., Agrawal, V. S., Marinella, M. J., Agarwal, S., & Bennett, C. H. (2026) · Nature Electronics · doi:10.1038/s41928-026-01725-x · Cited by 2
Self-citationSigning Crossbar Columns Instead of Neural Network Weights in Compute-in-Memory Accelerators
Pei, J., Wang, S., Yang, Y., Fang, J., 张久献, Shen, Y., Du, Y., Wang, Y., & Yang, Y. (2026) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1145/3770743.3804230
Quantized Analog Inference Hardware Enabled by MoS 2– Hf 1‐x Zr x O 2 FeFET Synapses
Park, J., Jung, M., Kim, H., Park, C., Lee, S., Kim, K. K., Yoon, H., Ahn, J., & Suh, D. (2026) · Advanced Science · doi:10.1002/advs.78366
“We then implemented a crossbar-level simulator in which synaptic weights were mapped directly onto the experimentally measured P/D conductance characteristics of the FeFET devices [ 46 ].”
Metadata
- Authors
- Xiao, T. P., Feinberg, B., Bennett, C. H., Prabhakar, V., Saxena, P., Agrawal, V., Agarwal, S., & Marinella, M. J.
- Affiliations
- Sandia National Laboratories California (US)
- Infineon Technologies (United States) (US)
- Arizona State University (US)
- ORCID
- Venue
- IEEE Circuits and Systems Magazine
- Published
- 2022
- Volume
- 22
- Issue
- 4
- Pages
- 26-48
- ISSN
- 0163-6812 (ISSN-L)
- 1531-636X
- 1558-0830
- OpenAlex
- W3198899759
- Cited by
- 76
- References
- 71
- Citations by year
- 2026: 18 · 2025: 30 · 2024: 18 · 2023: 8 · 2022: 2
- FWCI
- 3.97
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Circuits and Systems Magazine (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University) (cc-by)
- Funders
- U.S. Department of Energy (DENA0003525, NA0003525)
- National Nuclear Security Administration (NA0003525, DENA-0003525)
- Sandia National Laboratories (NA0003525, DE‐NA,0003525)
- 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 › Engineering › Electrical and Electronic Engineering › Analog and Mixed-Signal Circuit Design
- Keywords
- in-sensor computing, inference accuracy, bit slicing, analog-to-digital converter resolution, parasitic resistance, coding errors, ResNet50, energy efficiency
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.