Pith. sign in

REVIEW 1 major objections 1 minor 39 references

Harnessing the Full Potential of RRAMs through Scalable and Distributed In-Memory Computing with Integrated Error Correction

T0 review · 1 major / 1 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Two-tier error correction in the MELISO+ framework cancels first-order RRAM programming errors algebraically and denoises the residual, cutting arithmetic errors by over 90% and letting low-precision RRAM arrays outperform high-precision…

desk verdict Clever error-cancellation scheme and a useful RRAM simulation framework, but the described implementation computes A^T x~ instead of A x~, so the headline error-reduction claims only hold for symmetric matrices. read the letter →

arxiv 2508.13298 v2 pith:EXH72CZD submitted 2025-08-18 cs.DC cs.ARcs.ETcs.PFcs.SYeess.SY

Huynh Q. N. Vo , Md Tawsif Rahman Chowdhury , Paritosh Ramanan , Murat Yildirim , Gozde Tutuncuoglu This is my paper · ORCID
classification cs.DCcs.ARcs.ETcs.PFcs.SYeess.SY
keywords in-memorycomputingRRAMresistiverandomaccessmemorymatrix-vectormultiplicationerrorcorrectiondistributedwrite-and-verifyanalogcrossbararrays
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper introduces MELISO+, a full-stack framework for matrix-vector multiplication on RRAM in-memory hardware, and claims it removes the two obstacles that have kept such hardware from wide use: device variability and small array sizes. Its two-tier error correction cancels first-order programming errors algebraically and then denoises what remains, reducing arithmetic errors by more than 90% across several RRAM material systems. With that correction, low-precision, low-latency devices such as TaOx-HfOx can match the accuracy of the high-precision benchmark device while consuming three to five orders of magnitude less energy and running about 100 times faster. A distributed backend partitions large matrices and vectors row-wise across a multi-array cluster with a virtualization layer, letting fixed-size crossbars compute on matrices larger than 65,000 by 65,000. If these claims hold, RRAM memory arrays become a practical, energy-efficient substrate for the large matrix workloads behind modern AI.

What carries the argument

The load-bearing object is the algebraic cancellation identity $p = \tilde{A}x + A\tilde{x} - \tilde{A}\tilde{x} = Ax(1-\epsilon_A\epsilon_x)$, which holds when programming errors are multiplicative and stable. The vector $x$ is first written as repeated rows of a matrix $X^\top$ so that the three required products can be generated on the same array; reading them out and combining them removes the first-order error. The second stage is a regularized least-squares denoiser $y(\lambda) = (I_n + \lambda L^\top L)^{-1}p$, where $L$ is the first-order difference matrix, used to attenuate the remaining second-order term. The scalability claim is carried by a distributed mechanism: the matrix and vector are partitioned row-wise into chunks for individual crossbar arrays, coordinated through the Message Passing Interface (MPI), and a virtualization layer zero-pads dimensions and serially maps sub-blocks onto fixed-size arrays.

What would settle it

Program a real RRAM crossbar with a known matrix and vector, then read the same programmed array several times without reprogramming and compute the three products needed for $p$; if the residual after forming $p$ does not fall by roughly 90% relative to a naive $Ax$ product, or if errors do not follow the multiplicative form $\tilde A = A(1+\epsilon_A)$, the central claim is falsified. A simulation that injects additive or state-dependent noise instead of multiplicative noise should show the same failure.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the dominant programming error of an RRAM crossbar can be cancelled in software without changing the device. If encoding $A$ and $x$ as conductances introduces multiplicative perturbations, $\tilde A = A(1+\epsilon_A)$ and $\tilde x = x(1+\epsilon_x)$, then computing the three products $\tilde Ax$, $A\tilde x$, and $\tilde A\tilde x$ and forming $p = \tilde Ax + A\tilde x - \tilde A\tilde x$ yields $p = Ax(1-\epsilon_A\epsilon_x)$; every first-order error term is gone, leaving only the product of two small errors. The paper then applies a regularized least-squares denoiser, $y(\lambda) = (I_n + \lambda L^\top L)^{-1}p$ with $L$ the first-order difference matrix and $\lambda = 10^{-12}$, to suppress that second-order residual. On the scaling side, the paper shows that a fixed array of crossbars, with row-wise partitioning governed by MPI and a virtualization layer that serializes and zero-pads sub-blocks, can carry out reliable matrix-vector multiplication at dimensions beyond $65{,}000 \times 65{,}000$ with roughly linear memory and runtime scaling.

Load-bearing premise

The error reduction depends on each RRAM programming error being a small percentage error that stays identical every time the encoded matrix or vector is reused; if device noise is nonlinear, changes between reads, or includes read-out and conversion noise, the cancellation in Equation (7) breaks down.

Editorial extensions

If this is right

  • Lower-precision RRAM devices such as TaOx-HfOx can reach the accuracy of high-precision benchmark devices in matrix-vector multiplication, with the paper reporting 3-5 orders of magnitude lower write energy and roughly two orders lower write latency.
  • A fixed physical configuration of RRAM crossbars can process matrices far larger than its own footprint, with the paper demonstrating reliable computation at matrix sizes up to at least 65,025 by 65,025.
  • Because the two-tier correction is formulated from the algebra of the MVM itself, the paper argues it transfers across device material systems and error models without per-device retuning.
  • Combining write-and-verify iteration with error correction lets accuracy stabilize at small iteration counts, so the correction overhead does not erase the device's speed advantage.
  • The distributed and virtualized scheme gives roughly linear scaling of memory and runtime with problem size, making RRAM-based MVM practical for high-dimensional workloads.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the cancellation identity assumes stable multiplicative errors, its 90% figure should not be expected to survive in the presence of additive read noise, analogue-to-digital conversion error, or drift between the three passes; those effects would require a third correction tier.
  • The regularized denoiser assumes the true output vector is smooth in its index order; for matrices whose outputs are not ordered smoothly, the second-stage gain is untested, so the error reduction may be smaller on arbitrary matrix layouts.
  • A direct extension would be to run the same two-tier scheme on error models with state-dependent or nonlinear programming errors; the gap between observed and claimed error reduction would quantify how much of the result depends on the multiplicative perturbation assumption.
  • The framework supplies a concrete systems argument for using cheap, low-precision RRAM crossbars in AI inference, where the workload is dominated by large matrix-vector products, rather than reserving in-memory computing for small demonstrations.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

1 major / 1 minor

Summary. The paper introduces MELISO+, a full-stack simulation and benchmarking framework for RRAM-based in-memory computing. Its two main contributions are a two-tier error-correction scheme (a first-order cancellation that combines three matrix-vector products, followed by a regularized least-squares denoising step) and a distributed/virtualized execution backend that reportedly scales to matrices over 65,000x65,000. The manuscript claims over 90% reduction of arithmetic errors, three-to-five orders of magnitude energy improvement, and a 100-fold latency reduction, and that lower-precision RRAM devices (e.g., TaOx-HfOx) can thus outperform higher-precision devices such as EpiRAM. These claims are evaluated through simulations on four device types and several SuiteSparse matrices.

Significance. If the central claims are correct, MELISO+ would be a significant contribution: the first-order cancellation is parameter-free under the stated multiplicative error model, the distributed/virtualized MVM approach addresses a real scalability bottleneck, and the comparison of multiple RRAM device types with a public benchmark pipeline could be broadly useful. The manuscript also ships a complete simulator with MPI support, which is a concrete strength. However, the error-cancellation claim is load-bearing and is not established for general matrices under the described circuit-level implementation, and the energy/latency comparisons omit significant overheads; these issues currently prevent the paper from fully supporting its headline claims.

major comments (1)
  1. [Sections 2.2 and 4.2] The energy and latency metrics Ew and Lw are write-only metrics, but the error-correction procedure requires three MVM passes (A tilde{x}, tilde{A} x, tilde{A} tilde{x}), the construction and row-wise readout of the n x n matrix X^T, and an additional denoising MVM with (I + lambda L^T L)^{-1}. None of these overheads, nor read/ADC costs, appear to be included in the reported Ew and Lw. As a result, the claimed three-to-five orders of magnitude energy reduction and 100-fold latency reduction relative to EpiRAM are not justified by the data presented. The comparison should be end-to-end, including all extra writes, reads, and arithmetic operations introduced by the correction scheme.
minor comments (1)
  1. [Section 2.3.2 / Fig. 5] The normalization factor for virtualization (the dashed lines in Fig. 5) is described only qualitatively. Please provide the exact formula used for the normalization and state it in the caption or Methods.

Circularity Check

1 steps flagged · score 4.0 of 10

The first-order cancellation is parameter-free and self-contained, but the reported >90% error reduction is partly an in-sample fit because the regularization parameter λ and write-and-verify iteration counts are tuned on the same benchmark matrices used for the headline results.

  1. fitted input called prediction [Section 2.1 (second-stage denoising; adjustable Write-and-Verify tuning); Methods Section 4.3, Eq. (10)]
    "λ∈ (0, 1) is a regularization parameter—we selected λ = 10−12 in MELISO+ since it produced the best result; ... As the performance of the adjustableWriteandVerify scheme depends on the number of iterations, we conducted experiments with iteration counts ranging from k = 0 to k = 20."

    Both λ and the write-and-verify iteration count k are selected after observing results on the same benchmark matrices (bcsstk02, Iperturb, and the scalability matrices) that are then used to report MELISO+'s error reduction and energy/latency improvements. The second-order denoising and iterative-programming contributions to the claimed 'over 90%' reduction are therefore in-sample optima, not parameter-free predictions. The first-order identity p = A~x + A x~ − A~x~ is self-contained, so the circularity is partial, but the headline quantitative claim is partly fitted rather than predicted.

full rationale

The algebraic cancellation in Eq. (7) is self-contained for the multiplicative error model of Eqs. (2)-(3): given the three products as defined, the combination v − y + u leaves only the εA εx term, with no input-dependent constant tuned to produce the cancellation. No load-bearing self-citation appears: reference [5] supplies the MELISO simulation platform, while the correction identity and the NeuroSim+ device models are independent of that citation. The circularity found is localized to the empirical claim: λ = 10^-12 is explicitly chosen 'since it produced the best result' on the same matrices whose errors are then reported, and the k values (e.g., k=2 for most devices on Iperturb, k=5 on bcsstk02, k=11 for Ag-aSi) are selected after scanning the reported benchmarks. Thus the second-order and iterative-programming portions of the 'over 90%' reduction are fitted results, not out-of-sample predictions. Separately, and not counted in the circularity score, the Section 2.1 circuit description of the A x~ term—writing X~^T with rows x~^T and feeding columns a_j of A—computes A^T x~ rather than A x~; if taken literally, the Eq. (7) cancellation holds only for symmetric matrices. This is a correctness/construction concern, not an equivalence-by-construction of the derivation with its inputs, so it does not raise the circularity score.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The error-correction claim depends on a small multiplicative error model, noiseless read-out of encoded values, and smoothness of the true result; the paper fits λ and write-and-verify iterations on the same benchmark matrices used for evaluation.

free parameters (3)
  • λ (regularization strength) = 1e-12
    Selected because it 'produced the best result' (Section 2.1); no sensitivity analysis or independent validation is provided.
  • Write-and-verify iteration count k = up to 20, optimal per device not stated
    Iterations are varied from 0 to 20, but Table 1 does not state which k produced the reported numbers, making the comparison potentially post hoc.
  • Error tolerance ε in write-and-verify = not reported
    Algorithms 1 and 2 require a tolerance ε that determines when programming stops; its value is never given and directly affects all error metrics.
assumptions (3)
  • domain assumption RRAM programming errors are small multiplicative perturbations, i.e., Ã = A(1+ε_A) and x̃ = x(1+ε_x) as in Equations (2) and (3).
    The cancellation identity and all error-reduction claims depend on this multiplicative model; nonlinear conductance changes violate it.
  • domain assumption Encoded vector values can be read out exactly with no read noise, and the exact matrix A and vector x can be supplied as inputs with no DAC or conversion error.
    The three-product scheme uses recorded x̃ and ideal A and x; any additional noise breaks the cancellation in Equation (7).
  • domain assumption The true matrix-vector product result b is smooth, meaning its first differences are small, justifying Tikhonov denoising with the first-order differential matrix L.
    The second-order correction applies (I + λL^T L)^{-1}p; if b is not smooth, this filter biases the result, and the paper does not check this assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Harnessing the Full Potential of RRAMs through Scalable and Distributed In-Memory Computing with Integrated Error Correction." pith.science (2026). https://pith.science/paper/EXH72CZD

@misc{pith2026250813298,
  author       = {Pith},
  title        = {Pith review of: Harnessing the Full Potential of RRAMs through Scalable and Distributed In-Memory Computing with Integrated Error Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EXH72CZD}},
  note         = {Machine review of arXiv:2508.13298}
}
abstract

Exponential growth in global computing demand is exacerbated due to the higher-energy requirements of conventional architectures, primarily due to energy-intensive data movement. In-memory computing with Resistive Random Access Memory (RRAM) addresses this by co-integrating memory and processing, but faces significant hurdles related to device-level non-idealities and poor scalability for large computing tasks. Here, we introduce MELISO+ (In-Memory Linear Solver), a full-stack, distributed framework for energy-efficient in-memory computing. MELISO+ proposes a novel two-tier error correction mechanism to mitigate device non-idealities and develops a distributed RRAM computing framework to enable matrix computations exceeding dimensions of $65,000\times65,000$. This approach reduces first- and second-order arithmetic errors due to device non-idealities by over $90\%$, enhances energy efficiency by three to five orders of magnitude, and decreases latency 100-fold. Hence, MELISO+ allows lower-precision RRAM devices to outperform high-precision device alternatives in accuracy, energy and latency metrics. By unifying algorithm-hardware co-design with scalable architecture, MELISO+ significantly advances sustainable, high-dimensional computing suitable for applications like large language models and generative AI.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 20 canonical work pages

  1. [1]

    Challenges6(1), 117–157 (2015) https://doi.org/10.3390/ 18 challe6010117

    Andrae, A.S.G., Edler, T.: On global electricity usage of communication technol- ogy: Trends to 2030. Challenges6(1), 117–157 (2015) https://doi.org/10.3390/ 18 challe6010117

    work page 2015

  2. [2]

    SIGPLAN Not

    Boroumand, A., Ghose, S., Kim, Y., Ausavarungnirun, R., Shiu, E., Thakur, R., Kim, D., Kuusela, A., Knies, A., Ranganathan, P., Mutlu, O.: Google workloads for consumer devices: Mitigating data movement bottlenecks. SIGPLAN Not. 53(2), 316–331 (2018) https://doi.org/10.1145/3296957.3173177

    arXiv 2018

  3. [3]

    ArXiv abs/2401.14428 (2024)

    Khan,A.A.,Lima,J.P.C.,Farzaneh,H.,Castrillón,J.:Thelandscapeofcompute- near-memory and compute-in-memory: A research and commercial overview. ArXiv abs/2401.14428 (2024)

    arXiv 2024

  4. [4]

    Nature 615(7954), 823–829 (2023) https://doi.org/10.1038/s41586-023-05759-5

    Rao, M., Tang, H., Wu, J., Song, W., Zhang, M., Yin, W., Zhuo, Y., Kiani, F., Chen, B., Jiang, X., Liu, H., Chen, H.-Y., Midya, R., Ye, F., Jiang, H., Wang, Z., Wu, M., Hu, M., Wang, H., Xia, Q., Ge, N., Li, J., Yang, J.J.: Thousands of conductance levels in memristors integrated on CMOS. Nature 615(7954), 823–829 (2023) https://doi.org/10.1038/s41586-023...

    doi:10.1038/s41586-023-05759-5 2023

  5. [5]

    In: 2024 International Conference on Neuromorphic Systems (ICONS), pp

    Rahman Chowdhury, M.T., Nguyen Vo, H.Q., Ramanan, P., Yildirim, M., Tutun- cuoglu, G.: The lynchpin of in-memory computing: A benchmarking framework for vector-matrix multiplication in rrams. In: 2024 International Conference on Neuromorphic Systems (ICONS), pp. 336–342 (2024). https://doi.org/10.1109/ ICONS62911.2024.00058

    arXiv 2024

  6. [6]

    Advanced Materials Technolo- gies 10(4), 2400585 (2025) https://doi.org/10.1002/admt.202400585 https://advanced.onlinelibrary.wiley.com/doi/pdf/10.1002/admt.202400585

    Lee, Y., Jeon, B., Cho, Y., Kim, J., Shim, W., Kim, S.: Recent progress in memrsitor array structures and solutions for sneak path current reduction. Advanced Materials Technolo- gies 10(4), 2400585 (2025) https://doi.org/10.1002/admt.202400585 https://advanced.onlinelibrary.wiley.com/doi/pdf/10.1002/admt.202400585

    work page doi:10.1002/admt.202400585 2025

  7. [7]

    Chemical Reviews (2025) https://doi.org/10.1021/acs.chemrev.4c00845

    Ielmini, D., Pedretti, G.: Resistive Switching Random-Access Memory (RRAM): Applications and Requirements for Memory and Computing. Chemical Reviews (2025) https://doi.org/10.1021/acs.chemrev.4c00845 . Publisher: American Chemical Society. Accessed 2025-06-25

    work page doi:10.1021/acs.chemrev.4c00845 2025

  8. [8]

    In: 2022 12th International Conference on Electrical and Computer Engineering (ICECE), pp

    Sumon, M.M.H., Ali, M.L., Sadi, M.S.: An efficient triple-error correction method for memristor-based memory systems. In: 2022 12th International Conference on Electrical and Computer Engineering (ICECE), pp. 380–383 (2022). https: //doi.org/10.1109/ICECE57408.2022.10088479

    arXiv 2022

Show all 39 references
  1. [9]

    In: Proceedings of the 59th ACM/IEEE Design Automa- tion Conference

    Crafton, B., Wan, Z., Spetalnick, S., Yoon, J.-H., Wu, W., Tokunaga, C., De, V., Raychowdhury, A.: Improving compute in-memory ecc reliability with successive correction. In: Proceedings of the 59th ACM/IEEE Design Automa- tion Conference. DAC ’22, pp. 745–750. Association for...

    2022

  2. [10]

    In: 2020 IEEE 38th VLSI Test Symposium (VTS), pp

    Das, A., Touba, N.A.: Selective checksum based on-line error correction for rram based matrix operations. In: 2020 IEEE 38th VLSI Test Symposium (VTS), pp. 1–6 (2020). https://doi.org/10.1109/VTS48691.2020.9107606

    2020

  3. [11]

    ACM Trans

    Liu, M., Xia, L., Wang, Y., Chakrabarty, K.: Algorithmic fault detection for rram-based matrix operations. ACM Trans. Des. Autom. Electron. Syst.25(3) (2020) https://doi.org/10.1145/3386360

    2020 doi

  4. [13]

    IEEE Transactions on Computers C-19(5), 398–402 (1970) https://doi.org/10.1109/ T-C.1970.222937

    Rao, T.R.N.: Biresidue error-correcting codes for computer arithmetic. IEEE Transactions on Computers C-19(5), 398–402 (1970) https://doi.org/10.1109/ T-C.1970.222937

    1970

  5. [14]

    1–3 (2021)

    Hu, Y., Cheng, K., Zhang, Z., Wang, R., Wang, Y., Huang, R.: Error correction schemeforreliablerram-basedin-memorycomputing.In:20215thIEEEElectron Devices Technology & Manufacturing Conference (EDTM), pp. 1–3 (2021). https: //doi.org/10.1109/EDTM50988.2021.9420944

    2021

  6. [15]

    In: 2024 IEEE European Test Symposium (ETS), pp

    Parrini, L., Soliman, T., Hettwer, B., Borrmann, J.M., Singh, S., Bende, A., Rana, V., Merchant, F., Wehn, N.: Error detection and correction codes for safe in-memory computations. In: 2024 IEEE European Test Symposium (ETS), pp. 1–4 (2024). https://doi.org/10.1109/ETS61313.20...

    2024

  7. [16]

    In: 2024 IEEE 30th International Symposium on On-Line Testing and Robust Sys- tem Design (IOLTS), pp

    Papavramidou, P., Ricavy, S., Mounet, C., Jahan, C., Castellani, N., Andrieu, F.: A synergistic fault tolerance framework for mbit 28nm embedded rram. In: 2024 IEEE 30th International Symposium on On-Line Testing and Robust Sys- tem Design (IOLTS), pp. 1–7 (2024). https://doi....

    2024

  8. [17]

    In: 2018 IEEE International Symposium on High Performance Computer Architecture (HPCA), pp

    Feinberg, B., Wang, S., Ipek, E.: Making memristive neural network accelerators reliable. In: 2018 IEEE International Symposium on High Performance Computer Architecture (HPCA), pp. 52–65 (2018). https://doi.org/10.1109/HPCA.2018. 00015

    2018 doi

  9. [18]

    In: 2010 43rd Annual IEEE/ACM International Symposium on Microarchitecture, pp

    Seong, N.H., Woo, D.H., Srinivasan, V., Rivers, J.A., Lee, H.-H.S.: Safer: Stuck-at-fault error recovery for memories. In: 2010 43rd Annual IEEE/ACM International Symposium on Microarchitecture, pp. 115–124 (2010). https://doi. org/10.1109/MICRO.2010.46

    2010 doi

  10. [19]

    In: 2017 54th ACM/EDAC/IEEE Design Automation Conference (DAC), pp

    Liu, C., Hu, M., Strachan, J.P., Li, H.: Rescuing memristor-based neuromorphic design with high defects. In: 2017 54th ACM/EDAC/IEEE Design Automation Conference (DAC), pp. 1–6 (2017). https://doi.org/10.1145/3061639.3062310 20

    2017

  11. [20]

    In: 2019 56th ACM/IEEE Design Automation Conference (DAC), pp

    He, Z., Lin, J., Ewetz, R., Yuan, J.-S., Fan, D.: Noise injection adaption: End- to-end reram crossbar non-ideal effect adaption for neural network mapping. In: 2019 56th ACM/IEEE Design Automation Conference (DAC), pp. 1–6 (2019)

    2019

  12. [21]

    In: 2021 58th ACM/IEEE Design Automation Conference (DAC), pp

    Chen, C.-Y., Chakrabarty, K.: Pruning of deep neural networks for fault-tolerant memristor-based accelerators. In: 2021 58th ACM/IEEE Design Automation Conference (DAC), pp. 889–894 (2021). https://doi.org/10.1109/DAC18074. 2021.9586269

    2021

  13. [22]

    IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems43(2), 573–585 (2024) https://doi.org/10.1109/ TCAD.2023.3313089

    Eldebiky, A., Zhang, G.L., Böcherer, G., Li, B., Schlichtmann, U.: Correctnet+: Dealing with hw non-idealities in in-memory-computing platforms by error sup- pression and compensation. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems43(2), 573–585 ...

    2024

  14. [23]

    IEEE Transactions on Electron Devices 66(1), 793–801 (2019) https://doi.org/10.1109/TED.2018.2882779

    Dong, Z., Zhou, Z., Li, Z., Liu, C., Huang, P., Liu, L., Liu, X., Kang, J.: Convo- lutional neural networks based on rram devices for image recognition and online learning tasks. IEEE Transactions on Electron Devices 66(1), 793–801 (2019) https://doi.org/10.1109/TED.2018.2882779

    2019

  15. [24]

    Proceedings of the National Academy of Sciences 116(10), 4123–4128 (2019) https://doi.org/10.1073/pnas.1815682116 https://www.pnas.org/doi/pdf/10.1073/pnas.1815682116

    Sun, Z., Pedretti, G., Ambrosi, E., Bricalli, A., Wang, W., Ielmini, D.: Solving matrix equations in one step with cross-point resis- tive arrays. Proceedings of the National Academy of Sciences 116(10), 4123–4128 (2019) https://doi.org/10.1073/pnas.1815682116 https://www.pnas...

    2019 doi

  16. [25]

    Nature Communications 10(1) (2019) https://doi.org/ 10.1038/s41467-019-13103-7

    Mahmoodi, M.R., Prezioso, M., Strukov, D.B.: Versatile stochastic dot product circuits based on nonvolatile memories for high performance neurocomputing and neurooptimization. Nature Communications 10(1) (2019) https://doi.org/ 10.1038/s41467-019-13103-7

    2019 doi

  17. [26]

    In: 2015 IEEE International Electron Devices Meeting (IEDM), pp

    Chen, B., Cai, F., Zhou, J., Ma, W., Sheridan, P., Lu, W.D.: Efficient in-memory computing architecture based on crossbar arrays. In: 2015 IEEE International Electron Devices Meeting (IEDM), pp. 17–511754 (2015). https://doi.org/10. 1109/IEDM.2015.7409720

    2015

  18. [27]

    Pedretti, G., Milo, V., Ambrogio, S., Carboni, R., Bianchi, S., Calderoni, A., Ramaswamy, N., Spinelli, A.S., Ielmini, D.: Stochastic learning in neuromorphic hardwareviaspiketimingdependentplasticitywithrramsynapses.IEEEJournal on Emerging and Selected Topics in Circuits and ...

    2018

  19. [28]

    Multimedia Tools and Applications (2024) https://doi.org/10.1007/s11042-024-20016-1

    Sengar, S.S., Hasan, A.B., Kumar, S., Carroll, F.: Generative artificial intelli- gence: A systematic review and applications. Multimedia Tools and Applications (2024) https://doi.org/10.1007/s11042-024-20016-1

    2024 doi

  20. [29]

    IEEE Electron Device Letters 37(8), 994–997 (2016) https://doi.org/10.1109/LED.2016.2582859

    Woo, J., Moon, K., Song, J., Lee, S., Kwak, M., Park, J., Hwang, H.: Improved 21 synaptic behavior under identical pulses using alox/hfo2 bilayer rram array for neuromorphic systems. IEEE Electron Device Letters 37(8), 994–997 (2016) https://doi.org/10.1109/LED.2016.2582859

    2016

  21. [30]

    In: 2018 IEEE Symposium on VLSI Technology, pp

    Wu, W., Wu, H., Gao, B., Yao, P., Zhang, X., Peng, X., Yu, S., Qian, H.: A methodology to improve linearity of analog rram for neuromorphic computing. In: 2018 IEEE Symposium on VLSI Technology, pp. 103–104 (2018). https://doi. org/10.1109/VLSIT.2018.8510690

    2018

  22. [31]

    Nature materials17(4), 335–340 (2018) https://doi.org/10.1038/s41563-017-0001-5

    Choi, S., Tan, S., Li, Z., Kim, Y., Choi, C., Chen, P., Yeon, H., Yu, S., Kim, J.: Sige epitaxial memory for neuromorphic computing with reproducible high performance based on engineered dislocations. Nature materials17(4), 335–340 (2018) https://doi.org/10.1038/s41563-017-000...

    2018 doi

  23. [32]

    In: 2017 IEEE International Electron Devices Meeting (IEDM), pp

    Chen, P.-Y., Peng, X., Yu, S.: Neurosim+: An integrated device-to-algorithm framework for benchmarking synaptic devices and array architectures. In: 2017 IEEE International Electron Devices Meeting (IEDM), pp. 6–11614 (2017). https: //doi.org/10.1109/IEDM.2017.8268337

    2017

  24. [33]

    IEEE Micro39(6), 54–63 (2019) https://doi.org/10.1109/MM.2019.2943047

    Yin, S., Kim, Y., Han, X., Barnaby, H., Yu, S., Luo, Y., He, W., Sun, X., Kim, J.-J., Seo, J.-s.: Monolithically integrated rram- and cmos-based in-memory com- puting optimizations for efficient deep learning. IEEE Micro39(6), 54–63 (2019) https://doi.org/10.1109/MM.2019.2943047

    2019

  25. [34]

    Nano Letters 10(4), 1297–1301 (2010) https://doi.org/10.1021/nl904092h https://doi.org/10.1021/nl904092h

    Jo, S.H., Chang, T., Ebong, I., Bhadviya, B.B., Mazumder, P., Lu, W.: Nanoscale memristor device as synapse in neuromorphic systems. Nano Letters 10(4), 1297–1301 (2010) https://doi.org/10.1021/nl904092h https://doi.org/10.1021/nl904092h. PMID: 20192230

    2010 doi

  26. [35]

    Journal of Open Source Software4(35), 1244 (2019) https://doi.org/10.21105/ joss.01244

    Kolodziej, S.P., Aznaveh, M., Bullock, M., David, J., Davis, T.A., Henderson, M., Hu, Y., Sandstrom, R.: The suitesparse matrix collection website interface. Journal of Open Source Software4(35), 1244 (2019) https://doi.org/10.21105/ joss.01244

    2019

  27. [36]

    In: Recent 22 Advances in Parallel Virtual Machine and Message Passing Interface: 11th Euro- pean PVM/MPI Users’ Group Meeting Budapest, Hungary, September 19-22,

    Gabriel, E., Fagg, G.E., Bosilca, G., Angskun, T., Dongarra, J.J., Squyres, J.M., Sahay, V., Kambadur, P., Barrett, B., Lumsdaine, A.,et al.: Open mpi: Goals, concept, and design of a next generation mpi implementation. In: Recent 22 Advances in Parallel Virtual Machine and Me...

  28. [37]

    Computing in Science & Engineering23(4), 47–54 (2021)

    Dalcin, L., Fang, Y.-L.L.: mpi4py: Status update after 12 years of development. Computing in Science & Engineering23(4), 47–54 (2021)

    2021

  29. [38]

    Connexions4, 1–25 (2013)

    Selesnick, I.: Least squares with examples in signal processing. Connexions4, 1–25 (2013)

    2013

  30. [39]

    In: Proceedings of the International Conference on Neuromorphic Systems

    Luo, Y., Peng, X., Yu, S.: Mlp+neurosimv3.0: Improving on-chip learning performance with device to algorithm optimizations. In: Proceedings of the International Conference on Neuromorphic Systems. ICONS ’19. Association for Computing Machinery, New York, NY, USA (2019). https:...

    2019

  31. [2004]

    97–104 (2004)

    Proceedings 11, pp. 97–104 (2004). Springer

    2004

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.