Pith. sign in

REVIEW 4 major objections 5 minor 56 references

Lightweight LIF-only SNN accelerator using differential time encoding

T0 review · 4 major / 5 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 By representing spikes as time differences, a multiplier-free LIF-only accelerator merges spike trains with a compare-and-subtract rule and reports over 99% MNIST accuracy at sub-millisecond latencies.

desk verdict A concrete multiplier-free SNN accelerator, but the key merger proof ignores overflow symbols, leaving the central correctness claim unverified as written. read the letter →

arxiv 2505.11252 v1 pith:UZAUC63T submitted 2025-05-16 cs.NE eess.SP

Daniel Windhager , Lothar Ratschbacher , Bernhard A. Moser , Michael Lunglmayr This is my paper · ORCID
classification cs.NEeess.SP
keywords spikingneuralnetworksLIF-onlydifferentialtimeencodinglearnedhardwareacceleratorFPGAASICMNIST
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

The paper aims to show that spiking neural networks keep their efficiency advantage on real hardware when spikes are represented as time differences instead of absolute times. The central result is a multiplier-free feedforward network built only from leaky-integrate-and-fire (LIF) neurons, in which synaptic inputs are added to neuron potentials and thresholding produces output spikes. Its key hardware primitive is a spike merger that, by comparing and subtracting leading time gaps, combines any number of differential-time spike trains into one, and a proof of this merge is given. With a learned encoder whose weights are restricted to $\{-1, 0, 1\}$, the design reports $99.03\%$ accuracy on MNIST, roughly $0.29\,\mathrm{ms}$ per inference on an Ultrascale+ FPGA, and roughly $0.17\,\mathrm{ms}$ on a 7-nm open-source ASIC flow, with no multipliers in either target. This matters because it suggests low-latency, low-resource SNN inference is practical on standard digital hardware.

What carries the argument

The load-bearing object is the spike merger element, a hardware block that implements Proposition 1. Given two differential-time spike trains, at each step it outputs the smaller of the two leading delta times and subtracts that value from the other train's leading delta (or refills it from the next stored spike), so the merged train is produced without ever converting to absolute time; a binary tree of these blocks merges any number of input trains. The second essential piece is the differential time format itself, where a spike is the time since the previous spike and the all-ones symbol is an overflow marker; its bit width is chosen by minimizing $b\sum_i \lceil d_i/(2^b-1)\rceil$, which for the MNIST network reaches a minimum at $b=2$. The neuron core uses a decay of $\beta = 0.5$ and threshold $\theta = 1$, so the LIF update is a shift, an add, and a comparison; because spikes are single bits, weighted sums are conditional additions, and restricting the learned encoder weights to $\{-1, 0, 1\}$ keeps the encoder multiplier-free as well.

What would settle it

Train the same feedforward LIF network with learned encoding on CIFAR-10 in two versions, one with encoder weights limited to $\{-1, 0, 1\}$ and one unrestricted, and compare test accuracy; if the unrestricted version beats the restricted one by several points (the paper reports only $55.15\%$ for the restricted case), the premise that the ternary restriction is 'almost negligible' would be false.

Watch

Extended reading notes

Core claim

The central claim is that a feedforward LIF-only SNN can be computed with additions and threshold comparisons alone without losing accuracy, provided spike trains are kept in differential time and merged by the simple rule of Proposition 1: emit the smallest leading delta and subtract it from the other trains' leads. The paper proves that this procedure yields the correct merged delta sequence, because the leading deltas of the input trains are absolute distances to a common time origin, and subtracting the emitted minimum merely shifts that origin. The design reaches $99.03\%$ test accuracy on MNIST with a feedforward LIF network and a $9\times 9$ learned encoding, with per-image inference time of about $0.29\,\mathrm{ms}$ on an Ultrascale+ FPGA and about $0.17\,\mathrm{ms}$ on a 7-nm ASIC implementation, both without multipliers. The paper also reports $55.15\%$ on CIFAR-10 with the same feedforward approach, and identifies $2$ bits as the optimal delta width for the MNIST encoder under its bit-cost measure.

Load-bearing premise

The central premise is that restricting the learned encoder's weights to $\{-1, 0, 1\}$ costs almost no accuracy, a claim the paper inherits from its earlier work rather than demonstrating here; if the gap turns out large on other data or training setups, the 'lightweight with high accuracy' claim loses its basis.

Editorial extensions

If this is right

  • Feedforward LIF-only SNNs can be implemented on FPGAs and ASICs without any DSP slices: spike-weighted sums are conditional adds, and the ternary encoder adds only additions and subtractions.
  • The differential time representation with an overflow symbol is never worse than absolute time in bit count and can be much better for evenly spaced spikes; a 2-bit width suffices for the MNIST encoder.
  • The Proposition 1 merger tree works for any number of input spike trains, so the same architecture can process event-camera or level-crossing ADC streams without converting spikes to absolute time.
  • Because each network layer is directly instantiated and runs on the same merged event stream, deeper and wider feedforward LIF networks can be added at the cost of more merger-tree LUTs and weight memory, not more multipliers.

Reading between the lines

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

  • A direct test of the central premise would be to train the same feedforward LIF model on CIFAR-10 with and without the $\{-1, 0, 1\}$ encoder-weight restriction; the reported $55.15\%$ is only the restricted result, so the 'almost negligible' gap is asserted, not shown, for this dataset.
  • The compare-and-subtract merger is timing-agnostic, so it could be applied directly to event-based sensor streams without the patch-serialization used for static images; this is an extension the paper does not pursue.
  • Because the ASIC synthesis used only small SRAM macros from the open-source flow, the reported $0.17\,\mathrm{ms}$ and $538\,\mathrm{MHz}$ are presented as bounds; optimized wider SRAMs would likely shift both, an effect the authors flag but do not quantify.
  • The bit-cost function $b\sum_i \lceil d_i/(2^b-1)\rceil$ could be recomputed at runtime to adapt the delta width per layer or input distribution, something the fixed-2-bit design does not do.
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

4 major / 5 minor

Summary. The paper proposes a hardware accelerator for feedforward networks of leaky integrate-and-fire (LIF) neurons whose spikes are represented by differential time intervals rather than absolute timestamps. The authors argue that differential time encoding with learned patch-based encoding is bit-efficient, prove (Proposition 1) that differential spike trains can be merged by repeatedly emitting the smallest leading delta and subtracting it from the other leading deltas, and present an FPGA/ASIC implementation with no multipliers. They report >99% accuracy on MNIST, roughly 0.29 ms FPGA inference and 0.17 ms ASIC inference using the ASAP7 PDK, and compare favorably with earlier SNN accelerators. The paper extends the authors' prior designs [4,5] with the merger proof, an explicit bit-width analysis, and synthesis results.

Significance. The central claims are attractive and, if fully substantiated, would make a useful contribution: a multiplier-free LIF-only inference engine, an open-source ASAP7 synthesis flow, and sub-millisecond latency on an external benchmark. Strengths include the concrete area/power/timing tables, the use of an open-source PDK and toolchain, and the stated goal of a self-contained encoding justification. However, the main correctness proof and the hardware description do not currently account for overflow symbols in the actual encoding, and several accuracy-related claims are deferred to companion papers; therefore the significance is conditional on a substantial revision.

major comments (4)
  1. [Section IV, Proposition 1; Section V, Fig. 4] The merger proposition is stated and proved only for ordered sets of absolute spike times, where each delta is a difference of two actual spike times. The encoding defined in Section III, however, contains an overflow symbol 2^b-1 that represents elapsed time without a spike, and Section VII deploys b=2 on MNIST streams whose inter-spike differences exceed 3 (Fig. 1). Under the stated merge rule, overflow-derived deltas are indistinguishable from real spikes. For example, with b=2, stream A=[3,1] (overflow at relative time 3, spike at time 4) and stream B=[2] (spike at time 2) merge to [2,1,1], which the layer controller would read as spikes at times 2, 3, and 4, inserting a false spike at time 3. Neither Fig. 4 nor Section V-B describes a spike/overflow flag, and Section V-B explicitly states that output spikes are triggered only at input-spike times. As written, the proof does not cover the actual encoded streams and the hardware correctness claim is unverified. The revision should either extend Proposition 1 to overflow symbols with a type bit that suppresses false spikes, or show that the deployed encoding avoids overflows (which the histogram contradicts), and should verify the corrected merger in RTL simulation.
  2. [Section III, Eqs. (2)-(5) and Fig. 2] The bit-width cost formula is inconsistent with the stated meaning of the overflow symbol. The text says a difference d_i with 2^b-1 <= d_i < 2^(b+1)-1 requires two symbols, but Eq. (2) uses ceil(d_i/(2^b-1)) symbols, which gives one symbol for d_i = 2^b-1. This matters because the claimed optimum b=2 (Fig. 2) is computed from that formula. In addition, the histogram in Fig. 1 comes from a single 400-128-10 MNIST network, so the optimal-bit-width conclusion is an empirical statement for that configuration, not a general property; the paper should say so explicitly and, ideally, show sensitivity to network size and dataset.
  3. [Section VII, learned encoding] The no-multiplier property of the accelerator depends on restricting the learned encoding weights to {-1,0,1}. The paper states that the resulting accuracy loss is almost negligible and refers to [5], but no supporting numbers appear in this manuscript. Because this restriction is load-bearing for the central lightweight-accuracy claim, the revision should report the unrestricted-weight baseline accuracy on the same networks and datasets, or reproduce the relevant comparison from [5] in sufficient detail.
  4. [Section VI and Table III] The ASIC row in Table III reports 99.03% accuracy and about 6000 inferences/s, but Section VI describes the ASIC results as informative bounds due to non-ideal SRAM placement and does not state where the 99.03% figure comes from. If this accuracy is the trained software model's accuracy rather than that of the synthesized RTL or a bit-accurate hardware simulation, that should be stated explicitly; given the overflow-handling question above, an RTL-level accuracy check is needed before the hardware can be claimed to achieve the reported accuracy.
minor comments (5)
  1. [Section IV, Proposition 1] The proof does not define a tie-breaking rule when the first elements of two input streams are equal; specify whether simultaneous spikes are emitted as consecutive zero-delay symbols and how that interacts with the merge tree.
  2. [Section III, Eq. (5)] The quantity K_i is used in Eq. (5) but is not explicitly defined before its first use; define it as the number of inter-spike differences equal to i in the histogram.
  3. [Section III, Fig. 1 caption] The caption says difference times but the figure is a histogram of differential times; clarify the bit width used and how overflow symbols would be represented in the histogram.
  4. [References] References [33] and [46] are the same paper (Lunglmayr et al., DEXA 2021) and should be merged to avoid duplicate citations.
  5. [Table III] The comparisons in Table III mix different technology nodes, power-measurement conventions, and accuracy reporting conditions; add a footnote or table note stating what is and is not directly comparable.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: central claims are externally benchmarked and the merge proof is self-contained.

full rationale

The paper's central results are empirical and benchmarked rather than derived by circular definitions. The MNIST 99.03% accuracy is a trained-network test-set result on an external benchmark, and the FPGA/ASIC latency, power, and area figures come from synthesis of the described RTL on the ZCU102 board and the ASAP7 PDK, both external to the network's fitted parameters. Proposition 1 is a direct proof from the definition of differential time encoding; it does not import the conclusion as an assumption. The bit-width choice b=2 is an explicit cost-function optimization over the histogram of the authors' own trained network, not a hidden prediction dressed up as a result. The self-citations that do exist, namely [4] for the LIF parameters (β=0.5, θ=1) and [5] for the learned encoding and the claim that restricting encoding weights to {-1,0,1} costs 'almost negligible' accuracy, are references to separate falsifiable prior work by the same authors rather than circular reductions; the MNIST result in this paper independently exercises the restricted-weight LIF setting and thus provides external evidence for those choices. I note separately that the Section IV merge proof abstracts away the overflow symbols introduced in Section III, which is a potential correctness gap in the hardware semantics, but that is a technical soundness concern rather than a circularity: the proof does not reduce to its input by construction, so it does not affect the circularity score.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claims rest on engineering choices (bit width, LIF parameters, ternary weight restriction) and on the representativeness of the open-source synthesis flow. No new physical entities are introduced. The bit-width and parameter choices are empirically grounded in the authors' prior work or in a single-network histogram, not derived from first principles.

free parameters (3)
  • Differential time encoding bit width b = 2
    Chosen by minimizing BS(b) in Eq. (5) for a single trained 400-128-10 MNIST network (Fig. 2). Not shown to be optimal for other networks or datasets.
  • LIF parameters beta and theta = 0.5, 1.0
    Adopted from the authors' prior work [4] with the claim that this does not reduce capability; no derivation or empirical comparison is given in this paper.
  • MNIST network topology and learned encoding patch size = 400-800-512-256-10, 9x9 patches
    Design choices that directly affect the reported accuracy and timing; not justified as optimal in this paper.
assumptions (5)
  • domain assumption The cost function BS(b)=b*sum(ceil(di/(2^b-1))) in Eq. (3) is a valid measure of encoding efficiency.
    The paper assumes counting bits is the right optimization criterion for comparing encodings; it ignores hardware overheads like overflow handling, multiplexing, and control logic.
  • domain assumption LIF neuron with beta=0.5 and theta=1 is sufficiently expressive for the tested tasks.
    Taken from the authors' previous work [4]; not re-derived here.
  • domain assumption Restricting learned encoding weights to {-1, 0, 1} has negligible accuracy cost.
    The paper states the difference is almost negligible and references [5]; no direct comparison is shown in this paper.
  • domain assumption The learned encoding can be trained end-to-end with standard backpropagation (surrogate gradients).
    Training procedure is not described in this paper; referenced as learned encoding from [5].
  • domain assumption OpenROAD/ASAP7 synthesis results are representative of a real 7nm implementation.
    The authors themselves call the results 'informative bounds' because SRAM blocks dominate and floorplanning is suboptimal (Section VI).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Lightweight LIF-only SNN accelerator using differential time encoding." pith.science (2026). https://pith.science/paper/UZAUC63T

@misc{pith2026250511252,
  author       = {Pith},
  title        = {Pith review of: Lightweight LIF-only SNN accelerator using differential time encoding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UZAUC63T}},
  note         = {Machine review of arXiv:2505.11252}
}
read the original abstract

Spiking Neural Networks (SNNs) offer a promising solution to the problem of increasing computational and energy requirements for modern Machine Learning (ML) applications. Due to their unique data representation choice of using spikes and spike trains, they mostly rely on additions and thresholding operations to achieve results approaching state-of-the-art (SOTA) Artificial Neural Networks (ANNs). This advantage is hindered by the fact that their temporal characteristic does not map well to already existing accelerator hardware like GPUs. Therefore, this work will introduce a hardware accelerator architecture capable of computing feedforward LIF-only SNNs, as well as an accompanying encoding method to efficiently encode already existing data into spike trains. Together, this leads to a design capable of >99% accuracy on the MNIST dataset, with ~0.29ms inference times on a Xilinx Ultrascale+ FPGA, as well as ~0.17ms on a custom ASIC using the open-source predictive 7nm ASAP7 PDK. Furthermore, this work will showcase the advantages of the previously presented differential time encoding for spikes, as well as provide proof that merging spikes from different synapses given in differential time encoding can be done efficiently in hardware.

Figures

Figures reproduced from arXiv: 2505.11252 by the authors.

Figure 1
Figure 1. Histogram for the difference times when using the test set of the [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overall number of bits BS required to encode the entire spike train based on chosen bit width b for the differential time. cumulative sums first and subsequently merging the sequences in absolute time. We describe the proof for two sequences, since it can trivially be extended to any number of sequences. Let (A = {ai} n i=1, ≤) be the ordered set of the absolute spike times of a spike train (the most commonly used o… view at source ↗
Figure 4
Figure 4. Spike merger element capable of merging two separate spike trains [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: Hardware architecture capable of accelerating arbitrarily deep and wide feedforward networks of LIF neurons [5]. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: ASIC layout using open-source ORFs and ASAP7 PDK with [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: 3x3 learned encoding by using patch-wise selection of the pixels with [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 38 canonical work pages

  1. [2]

    Are snns really more energy-efficient than anns? an in-depth hardware- aware study,

    M. Dampfhoffer, T. Mesquida, A. Valentian, and L. Anghel, “Are snns really more energy-efficient than anns? an in-depth hardware- aware study,” IEEE Transactions on Emerging Topics in Computational Intelligence, vol. 7, no. 3, pp. 731–741, 2023

    work page 2023

  2. [5]

    Spiking Neural Network Accelerator Architecture for Differential-Time Representation using Learned Encoding

    D. Windhager, L. Ratschbacher, B. A. Moser, and M. Lunglmayr, “Spiking neural network accelerator architecture for differential-time representation using learned encoding,” accepted for publication in Proc. of ISCAS2025. [Online]. Available: https://arxiv.org/abs/2501.07952

    work page Pith review arXiv

  3. [1]

    Gpt-4 technical report,

    OpenAI, J. Achiam, S. Adler, S. Agarwal et al. , “Gpt-4 technical report,” 2024. [Online]. Available: https://arxiv.org/abs/2303.08774

    arXiv 2024

  4. [3]

    On the sampling sparsity of analog- to-spike conversion based on leaky integrate-and-fire,

    B. A. Moser and M. Lunglmayr, “On the sampling sparsity of analog- to-spike conversion based on leaky integrate-and-fire,” Neuromorphic Computing and Engineering , vol. 5, no. 1, p. 014011, mar 2025. [Online]. Available: https://dx.doi.org/10.1088/2634-4386/adb884

    work page doi:10.1088/2634-4386/adb884 2025

  5. [4]

    Snn architecture for differential time encoding using decoupled processing time,

    D. Windhager, B. A. Moser, and M. Lunglmayr, “Snn architecture for differential time encoding using decoupled processing time,” 2023. [Online]. Available: https://arxiv.org/abs/2311.14447

    arXiv 2023

  6. [6]

    MNIST Handwritten Digit Database,

    Y . LeCun, C. Cortes, and C. Burges, “MNIST Handwritten Digit Database,” 2010. [Online]. Available: http://yann.lecun.com/exdb/mnist

    2010

  7. [7]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, “Learning multiple layers of features from tiny images,”

  8. [8]

    Asap7: A 7-nm finfet predictive process design kit,

    L. T. Clark, V . Vashishtha, L. Shifren, A. Gujja, S. Sinha, B. Cline, C. Ramamurthy, and G. Yeric, “Asap7: A 7-nm finfet predictive process design kit,” Microelectronics Journal , vol. 53, pp. 105–115, 2016. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ S002626921630026X

    work page 2016

Show all 56 references
  1. [9]

    Comprehensive online training and deployment for spiking neural networks,

    Z. Hao, Y . Huang, Z. Xu, Z. Yu, and T. Huang, “Comprehensive online training and deployment for spiking neural networks,” 2024. [Online]. Available: https://arxiv.org/abs/2410.07547

    2024

  2. [10]

    Enlarge: An efficient snn simulation framework on gpu clusters,

    P. Qu, H. Lin, M. Pang, X. Liu, W. Zheng, and Y . Zhang, “Enlarge: An efficient snn simulation framework on gpu clusters,” IEEE Transactions on Parallel and Distributed Systems , vol. 34, no. 9, pp. 2529–2540, 2023

    2023

  3. [11]

    A gpu based simulation of multilayer spiking neural networks,

    A. Ahmadi and H. Soleimani, “A gpu based simulation of multilayer spiking neural networks,” in 2011 19th Iranian Conference on Electrical Engineering, 2011, pp. 1–1

    2011

  4. [12]

    Towards scalable gpu-accelerated snn training via temporal fusion,

    Y . Li, J. Li, K. Sun, L. Leng, and R. Cheng, “Towards scalable gpu-accelerated snn training via temporal fusion,” 2024. [Online]. Available: https://arxiv.org/abs/2408.00280

    2024 arXiv

  5. [13]

    Gpu-ranc: A cuda accelerated simulation framework for neuromorphic architectures,

    S. Hassan, M. Inouye, M. C. Gonzalez, I. Aliyev, J. Mack, M. Hafiz, and A. Akoglu, “Gpu-ranc: A cuda accelerated simulation framework for neuromorphic architectures,” 2024. [Online]. Available: https://arxiv.org/abs/2404.16208

    2024 arXiv

  6. [14]

    Loihi: A neuromorphic manycore processor with on-chip learning,

    M. Davies, N. Srinivasa, T.-H. Lin, G. Chinya, Y . Cao, S. H. Choday, G. Dimou, P. Joshi, N. Imam, S. Jain, Y . Liao, C.-K. Lin, A. Lines, R. Liu, D. Mathaikutty, S. McCoy, A. Paul, J. Tse, G. Venkataramanan, Y .-H. Weng, A. Wild, Y . Yang, and H. Wang, “Loihi: A neuromorphic ...

    2018

  7. [15]

    The spinnaker project,

    S. B. Furber, F. Galluppi, S. Temple, and L. A. Plana, “The spinnaker project,” Proceedings of the IEEE , vol. 102, no. 5, pp. 652–665, 2014

    2014

  8. [16]

    Spinnaker 2: A 10 million core processor system for brain simulation and machine learning,

    C. Mayr, S. Hoeppner, and S. Furber, “Spinnaker 2: A 10 million core processor system for brain simulation and machine learning,” 2019. [Online]. Available: https://arxiv.org/abs/1911.02385

    2019 arXiv

  9. [17]

    Sparrowsnn: A hardware/software co-design for energy efficient ecg classification,

    Z. Yan, Z. Bai, T. Mitra, and W.-F. Wong, “Sparrowsnn: A hardware/software co-design for energy efficient ecg classification,”

  10. [18]

    The impact of the mit-bih arrhythmia database,

    G. Moody and R. Mark, “The impact of the mit-bih arrhythmia database,” IEEE Engineering in Medicine and Biology Magazine , vol. 20, no. 3, pp. 45–50, 2001

    2001

  11. [19]

    Energy-efficient high- accuracy spiking neural network inference using time-domain neurons,

    J. Song, J. Shin, H. Kim, and W.-S. Choi, “Energy-efficient high- accuracy spiking neural network inference using time-domain neurons,”

  12. [20]

    A hybrid ann-snn architecture for low-power and low-latency visual perception,

    A. Aydin, M. Gehrig, D. Gehrig, and D. Scaramuzza, “A hybrid ann-snn architecture for low-power and low-latency visual perception,”

  13. [21]

    Low-power spiking neural network audio source localisation using a hilbert transform audio event encoding scheme,

    S. Haghighatshoar and D. R. Muir, “Low-power spiking neural network audio source localisation using a hilbert transform audio event encoding scheme,” Communications Engineering , vol. 4, no. 1, Feb. 2025. [Online]. Available: http://dx.doi.org/10.1038/s44172-025-00359-9

    2025 doi

  14. [22]

    Energy-aware fpga implementation of spiking neural network with lif neurons,

    A. H. Ali, M. Navardi, and T. Mohsenin, “Energy-aware fpga implementation of spiking neural network with lif neurons,” 2024. [Online]. Available: https://arxiv.org/abs/2411.01628

    2024 arXiv

  15. [23]

    Deepfire2: A convolutional spiking neural network accelerator on fpgas,

    M. T. L. Aung, D. Gerlinghoff, C. Qu, L. Yang, T. Huang, R. S. M. Goh, T. Luo, and W.-F. Wong, “Deepfire2: A convolutional spiking neural network accelerator on fpgas,” IEEE Transactions on Computers, vol. 72, no. 10, p. 2847–2857, Oct. 2023. [Online]. Available: http://dx.doi...

    2023

  16. [24]

    Available: https://arxiv.org/abs/2303.14176

    [Online]. Available: https://arxiv.org/abs/2303.14176

    arXiv

  17. [25]

    Firefly: A high- throughput hardware accelerator for spiking neural networks with effi- cient dsp and memory optimization,

    J. Li, G. Shen, D. Zhao, Q. Zhang, and Y . Zeng, “Firefly: A high- throughput hardware accelerator for spiking neural networks with effi- cient dsp and memory optimization,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems , vol. 31, no. 8, pp. 1178–1191, 2023

    2023

  18. [26]

    Syncnn: Evaluating and accel- erating spiking neural networks on fpgas,

    S. Panchapakesan, Z. Fang, and J. Li, “Syncnn: Evaluating and accel- erating spiking neural networks on fpgas,” in 2021 31st International Conference on Field-Programmable Logic and Applications (FPL) , 2021, pp. 286–293

    2021

  19. [27]

    S2n2: A fpga accelerator for streaming spiking neural networks,

    A. Khodamoradi, K. Denolf, and R. Kastner, “S2n2: A fpga accelerator for streaming spiking neural networks,” in The 2021 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays, ser. FPGA ’21. New York, NY , USA: Association for Computing Machinery, 2021, p. 194–20...

    2021

  20. [28]

    A low power and low latency fpga-based spiking neural network accelerator,

    H. Liu, Y . Chen, Z. Zeng, M. Zhang, and H. Qu, “A low power and low latency fpga-based spiking neural network accelerator,” in 2023 International Joint Conference on Neural Networks (IJCNN) , 2023, pp. 1–8

    2023

  21. [29]

    Unsupervised aer object recognition based on multiscale spatio-temporal features and spiking neurons,

    Q. Liu, G. Pan, H. Ruan, D. Xing, Q. Xu, and H. Tang, “Unsupervised aer object recognition based on multiscale spatio-temporal features and spiking neurons,” IEEE Transactions on Neural Networks and Learning Systems, vol. 31, no. 12, pp. 5300–5311, 2020

    2020

  22. [30]

    Impact of the aer- induced timing distortion on spiking neural networks implementing dsp,

    T. Mesquida, A. Valentian, D. Bol, and E. Beigne, “Impact of the aer- induced timing distortion on spiking neural networks implementing dsp,” in 2016 12th Conference on Ph.D. Research in Microelectronics and Electronics (PRIME), 2016, pp. 1–4

    2016

  23. [31]

    Efficient hardware implementa- tion of stdp for aer based large-scale snn neuromorphic system,

    J. Kim, J. Park, S. Joo, and S.-O. Jung, “Efficient hardware implementa- tion of stdp for aer based large-scale snn neuromorphic system,” in 2020 35th International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC) , 2020, pp. 1–4

    2020

  24. [32]

    Ac- celerating spike-by-spike neural networks on fpga with hybrid custom floating-point and logarithmic dot-product approximation,

    Y . Nevarez, D. Rotermund, K. R. Pawelzik, and A. Garcia-Ortiz, “Ac- celerating spike-by-spike neural networks on fpga with hybrid custom floating-point and logarithmic dot-product approximation,” IEEE Access, vol. 9, pp. 80 603–80 620, 2021

    2021

  25. [33]

    Robust and efficient bio- inspired data-sampling prototype for time-series analysis,

    M. Lunglmayr, G. Lindorfer, and B. Moser, “Robust and efficient bio- inspired data-sampling prototype for time-series analysis,” in DEXA

  26. [34]

    Fast and low-power leading-one detectors for energy-efficient logarithmic computing,

    M. S. Ansari, S. Gandhi, B. F. Cockburn, and J. Han, “Fast and low-power leading-one detectors for energy-efficient logarithmic computing,” IET Computers & Digital Techniques , vol. 15, no. 4, pp. 241–250, 2021. [Online]. Available: https://ietresearch.onlinelibrary. wiley.com...

    2021 doi

  27. [35]

    Leading one detectors and leading one position detectors - an evolutionary design methodology,

    K. Kunaraj and R. Seshasayanan, “Leading one detectors and leading one position detectors - an evolutionary design methodology,” Canadian Journal of Electrical and Computer Engineering, vol. 36, no. 3, pp. 103– 110, 2013

    2013

  28. [36]

    Neuromorphic computing with aer using time-to-event-margin propagation,

    M. S. R, S. Chakrabartty, and C. S. Thakur, “Neuromorphic computing with aer using time-to-event-margin propagation,” 2023. [Online]. Available: https://arxiv.org/abs/2304.13918

    2023 arXiv

  29. [37]

    Approximate leading one detector design for a hardware-efficient mitchell multiplier,

    S. Gandhi, M. S. Ansari, B. F. Cockburn, and J. Han, “Approximate leading one detector design for a hardware-efficient mitchell multiplier,” in 2019 IEEE Canadian Conference of Electrical and Computer Engi- neering (CCECE), 2019, pp. 1–4

    2019

  30. [38]

    Toward an open-source digital flow: First learnings from the openroad project,

    T. Ajayi, V . A. Chhabria, M. Fogac ¸a, S. Hashemi, A. Hosny, A. B. Kahng, M. Kim, J. Lee, U. Mallappa, M. Neseem et al. , “Toward an open-source digital flow: First learnings from the openroad project,” in Proceedings of the 56th Annual Design Automation Conference 2019 , 201...

    2019

  31. [39]

    On the versatility of the ihp bicmos open source and manufacturable pdk: A step towards the future where anybody can design and build a chip,

    K. Herman, N. Herfurth, T. Henkes, S. Andreev, R. Scholz, M. M ¨uller, M. Krattenmacher, H. Pretl, and W. Grabinski, “On the versatility of the ihp bicmos open source and manufacturable pdk: A step towards the future where anybody can design and build a chip,” IEEE Solid- Stat...

    2024

  32. [40]

    GitHub repository of the IIC-OSIC- TOOLS,

    H. Pretl and G. Zachl, “GitHub repository of the IIC-OSIC- TOOLS,” Sep. 2023. [Online]. Available: https://github.com/iic-jku/ IIC-OSIC-TOOLS

    2023

  33. [41]

    Vlsi implementations of low-power leading-one detector circuits,

    K. Abed and R. Siferd, “Vlsi implementations of low-power leading-one detector circuits,” in Proceedings of the IEEE SoutheastCon 2006, 2006, pp. 279–284

    2006

  34. [42]

    A 4096-neuron 1m-synapse 3.8-pj/sop spiking neural network with on-chip stdp learning and sparse weights in 10-nm finfet cmos,

    G. K. Chen, R. Kumar, H. E. Sumbul, P. C. Knag, and R. K. Krishna- murthy, “A 4096-neuron 1m-synapse 3.8-pj/sop spiking neural network with on-chip stdp learning and sparse weights in 10-nm finfet cmos,” IEEE Journal of Solid-State Circuits, vol. 54, no. 4, pp. 992–1002, 2019

    2019

  35. [43]

    Scalable energy-efficient, low-latency implementations of trained spik- ing deep belief networks on spinnaker,

    E. Stromatias, D. Neil, F. Galluppi, M. Pfeiffer, S.-C. Liu, and S. Furber, “Scalable energy-efficient, low-latency implementations of trained spik- ing deep belief networks on spinnaker,” in 2015 International Joint Conference on Neural Networks (IJCNN) , 2015, pp. 1–8

    2015

  36. [44]

    Always-on sub-microwatt spiking neural network based on spike-driven clock- and power-gating for an ultra-low-power intelligent device,

    P. K. Chundi, D. Wang, S. J. Kim, M. Yang, J. P. Cerqueira, J. Kang, S. Jung, S. Kim, and M. Seok, “Always-on sub-microwatt spiking neural network based on spike-driven clock- and power-gating for an ultra-low-power intelligent device,” Frontiers in Neuroscience , vol. V olume...

    2021

  37. [45]

    Encoding, model, and architecture: Systematic optimization for spiking neural network in fpgas,

    H. Fang, Z. Mei, A. Shrestha, Z. Zhao, Y . Li, and Q. Qiu, “Encoding, model, and architecture: Systematic optimization for spiking neural network in fpgas,” in 2020 IEEE/ACM International Conference On Computer Aided Design (ICCAD) , 2020, pp. 1–9

    2020

  38. [46]

    Robust 7-nm sram design on a predictive pdk,

    V . Vashishtha, M. Vangala, P. Sharma, and L. T. Clark, “Robust 7-nm sram design on a predictive pdk,” in2017 IEEE International Symposium on Circuits and Systems (ISCAS) , 2017, pp. 1–4. IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS 9

    2017

  39. [47]

    A 61-nw level-crossing adc with adaptive sampling for biomedical applications,

    Y . Hou, J. Qu, Z. Tian, M. Atef, K. Yousef, Y . Lian, and G. Wang, “A 61-nw level-crossing adc with adaptive sampling for biomedical applications,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 66, no. 1, pp. 56–60, 2019

    2019

  40. [48]

    A fixed window level crossing adc with activity dependent power dissipation,

    A. Ogweno, P. Degenaar, V . Khomenko, and A. Yakovlev, “A fixed window level crossing adc with activity dependent power dissipation,” in 2016 14th IEEE International New Circuits and Systems Conference (NEWCAS), 2016, pp. 1–4

    2016

  41. [49]

    Training spiking neural networks using lessons from deep learning,

    J. K. Eshraghian, M. Ward, E. Neftci, X. Wang, G. Lenz, G. Dwivedi, M. Bennamoun, D. S. Jeong, and W. D. Lu, “Training spiking neural networks using lessons from deep learning,” Proceedings of the IEEE , vol. 111, no. 9, pp. 1016–1054, 2023

    2023

  42. [50]

    Cifar-10 analysis with a neural net- work,

    V . Krikonis, “Cifar-10 analysis with a neural net- work,” https://www.kaggle.com/code/vassiliskrikonis/ cifar-10-analysis-with-a-neural-network, 2021, accessed: 2025-04- 28

    2021

  43. [51]

    Robust and efficient bio- inspired data-sampling prototype for time-series analysis,

    M. Lunglmayr, G. Lindorfer, and B. Moser, “Robust and efficient bio- inspired data-sampling prototype for time-series analysis,” in Database and Expert Systems Applications - DEXA 2021 Workshops . Cham: Springer International Publishing, 2021, pp. 119–126

    2021

  44. [56]

    Cifar-10 classification using linear models, ann, and cnn,

    A. Jahantab, “Cifar-10 classification using linear models, ann, and cnn,” https://github.com/amirjahantab/CIFAR-10, 2023, accessed: 2025-04-28

    2023

  45. [2009]

    Available: https://api.semanticscholar.org/CorpusID: 18268744

    [Online]. Available: https://api.semanticscholar.org/CorpusID: 18268744

  46. [2021]

    Springer International Publishing, 2021, pp. 119–126

    2021

  47. [2022]

    Available: https://arxiv.org/abs/2202.02015

    [Online]. Available: https://arxiv.org/abs/2202.02015

    arXiv

  48. [2024]

    Available: https://arxiv.org/abs/2406.06543

    [Online]. Available: https://arxiv.org/abs/2406.06543

    arXiv

Pith tools

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