Pith. sign in

REVIEW 3 major objections 4 minor 42 references

Efficient Spatio-Temporal Signal Recognition on Edge Devices Using PointLCA-Net

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read PointLCA-Net claims 93.41% on DVS128 with about 92% lower estimated energy than point-cloud SNNs.

desk verdict Clean proof of concept, but the 92% energy claim is not like-for-like because it excludes the feature extractor. read the letter →

arxiv 2411.14585 v3 pith:XDC6623V submitted 2024-11-21 cs.LG cs.ET

Sanaz Mahmoodi Takaghaj This is my paper · ORCID
classification cs.LGcs.ET
keywords ExemplarLCAPointNetsspatio-temporalsignalsneuromorphiccomputingevent-basedvisionenergyefficiencyDVS128sparsecoding
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

PointLCA-Net is a proof of concept that event-camera and cochlea-derived spatio-temporal signals, converted to (x, y, t) point clouds, can be recognized without training a deep spiking network. The paper's recipe is two-stage: a pre-trained PointNet or PointNet++ extracts a 1024-dimensional feature vector from each event cloud, and a single-layer Exemplar LCA encoder-decoder classifies those frozen features using sparse, competitive neuron dynamics. That matters because edge devices run under tight power and memory budgets, so shifting the cost into one-time feature extraction and sparse inference is an attractive trade. The strongest reported result is 93.41% top-1 accuracy on the 11-class DVS128 gesture benchmark with PointNet++ features, at an estimated 0.065 mJ for the LCA stage—the basis for the claimed roughly 92% energy reduction compared with other spiking point-cloud networks. The paper is explicit that the energy figures are theoretical and that the PointNet feature-extraction stage is excluded from the FLOP and joule accounting, so the headline saving applies to the LCA decoder rather than the whole pipeline.

What carries the argument

The central object is the Exemplar LCA-Decoder, a single-layer spiking encoder-decoder whose dictionary columns are the PointNet feature vectors $\varphi_i$. Each LIF neuron integrates an excitatory input $b_i = S\varphi_i$ and is suppressed by other active neurons through the Gramian matrix $G = \varphi^T\varphi$; a threshold $\lambda$ in Eq. (6) keeps the activation sparse, and the paper reports that about 0.4% of neurons are active at each time step. Classification uses either the maximum activation or the maximum sum of activations per class, so no trained readout layer is required. This machinery turns the deep PointNet extractor into a one-time stored dictionary and makes the per-test computation a crossbar-friendly vector-matrix product with sparse neuron activity.

What would settle it

Compute the FLOP count of the PointNet/PointNet++ feature extractor on the 1024-event DVS128 point clouds and add it to the 0.7 GFLOPs attributed to the LCA decoder; if that added cost is comparable to or larger than the decoder cost, the claimed 92% energy reduction over SpikePoint does not hold for an end-to-end deployment.

Watch

Extended reading notes

Core claim

The paper's central claim is that a dictionary of PointNet/PointNet++ features can serve as the synaptic memory of an Exemplar LCA encoder-decoder, making spatio-temporal event recognition both accurate and sparse. Training consists of extracting features once from the training point clouds and storing each feature vector as a column of a memristor crossbar; no dictionary learning or decoder backpropagation is needed. At inference, LCA dynamics—excitatory drive, Gramian inhibition, and thresholding—select a tiny fraction of active neurons, and decoding by maximum sum of per-class activations yields 98.78% on NMNIST, 78.46% on SHD, and 93.41% on DVS128 with PointNet++ features. The paper further claims that LCA sparsity cuts inference computation by an average of 99.54%, and that mapping the crossbar to an RRAM array at $9.09\times10^{-14}$ J/FLOP gives an estimated 0.065 mJ for DVS128 inference, about 24 times lower than the reported SpikePoint energy.

Load-bearing premise

The load-bearing premise is that the FLOP and energy estimates for the LCA decoder—the 0.4% active-neuron fraction and the $9.09\times10^{-14}$ J/FLOP RRAM figure—represent the deployed system's real cost, even though the paper explicitly excludes the PointNet feature-extraction stage from all accounting.

Editorial extensions

If this is right

  • Across three event-based datasets (NMNIST, DVS128, SHD), the same two-stage pipeline achieves high top-1 accuracy without any backpropagation in the spiking stage.
  • LCA sparsity reduces average inference computation by 99.54%, and cutting the integration interval from 100 to 10 time steps lowers inference FLOPs by roughly 80%.
  • With PointNet++ features and the maximum-sum decoder, DVS128 accuracy reaches 93.41% at an estimated 0.7 GFLOPs and 0.065 mJ per inference, giving the claimed order-of-magnitude energy advantage over the point-cloud SNN baseline.
  • Because the dictionary is assembled from stored features rather than learned, the same algorithm can be applied uniformly to a new spatio-temporal dataset by changing only the feature-extraction front end.

Reading between the lines

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

  • The paper leaves implicit that the ~92% energy saving is an LCA-stage figure: adding the excluded PointNet/PointNet++ FLOPs would determine whether the saving survives at the system level, and that calculation is an immediate next step.
  • The DVS128 comparison mixes protocols—PointLCA-Net uses all 11 classes with a fixed 28,606/7,408 split, while the cited 90.20% PointNet baseline used 10 classes—so a matched 10-class run would isolate how much of the gain is architectural rather than evaluational.
  • The paper sketches dictionary compression (PCA, SVM, discriminative feature selection) as future work; since the dictionary holds every training point, a natural stress test is to measure accuracy and FLOPs as the dictionary is pruned on a larger continuous event stream, where the full-dictionary approach would not scale.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes PointLCA-Net, a two-stage pipeline for spatio-temporal signal recognition on edge devices: a pre-trained PointNet/PointNet++ extracts a 1024-dimensional global feature vector from event-based point clouds, and an Exemplar LCA encoder-decoder then performs sparse coding and classification using a dictionary of stored training features. The method is evaluated on NMNIST, DVS128, and SHD, reporting top-1 accuracies up to 98.78%, 93.41%, and 78.46% respectively. The central claim is that PointLCA-Net achieves high accuracy while reducing energy consumption by approximately 92% compared to other spiking neural networks applied to point clouds, based on FLOP estimates and an RRAM crossbar energy model.

Significance. The idea of decoupling feature extraction from a neuromorphic sparse-coding classifier is conceptually attractive and the paper is, to its credit, explicit about the experimental setup, hyperparameters, and the fact that energy numbers are theoretical estimates. The work also extends the event-cloud formulation to NMNIST and SHD for the first time, which is a useful data point. The accuracy results are on external benchmarks and the core LCA-based classification pipeline is reproducible in principle from the equations given. However, the headline energy-efficiency claim depends on an incomplete accounting of the deployed system, so the significance of the paper as a hardware-energy contribution cannot be assessed until that accounting is corrected.

major comments (3)
  1. [IV.E, Table II] The energy comparison against SpikePoint is not like-for-like. Section IV.E explicitly states that the FLOP estimates 'exclude the FLOPs required for feature extraction by PointNet and PointNet++,' yet Table II compares the resulting 0.065 mJ for DVS128 against SpikePoint's full-pipeline dynamic (0.82 mJ) and static (0.756 mJ) energy. The abstract's claim of 'reducing energy consumption by approximately 92%' is therefore not supported for the full system. The feature-extraction stage (PointNet with input and feature T-Nets, MLP(64,64,64,128,1024) over 1024 points) performs a substantial number of MACs that must either be added to the PointLCA-Net total or explicitly excluded from the comparison with a clear statement that the comparison is LCA-stage-only. As written, this is a load-bearing omission for the central energy-efficiency claim.
  2. [IV.E, Eq. (10), Table I] The sparsity estimate \hat{M}, which drives the inference FLOP count, is inherited from reference [28] with the statement 'we have verified that this estimate is accurate.' No verification data for the three datasets used here (NMNIST, DVS128, SHD) is provided. Since the '99.54% reduction in computational effort' and all derived energy numbers scale linearly with \hat{M}, the paper should report measured active-neuron fractions for each dataset and each feature extractor (PointNet vs PointNet++). If the fraction differs from 0.4%, the workload and energy conclusions change proportionally.
  3. [IV.F, Section V] The energy per FLOP of 9.09e-14 J is taken from Yao et al. [37], which was measured for MAC operations in a memristor CNN accelerator. PointLCA-Net's LCA decoder includes thresholding (Eq. 6), membrane potential updates (Eq. 2), and peripheral input/output circuits (Section IV.F), not only MACs. Applying the same per-FLOP figure to all these operations is an unvalidated extrapolation. The paper should either model the hardware mapping more carefully (e.g., distinguishing analog MAC energy from digital control/peripheral energy) or clearly state that the reported energy is a MAC-only lower bound. Without this, the 92% reduction claim is not robust.
minor comments (4)
  1. [IV.E] The equation is labeled 'F LOP s(Inf erenec)' — the word 'Inference' is misspelled; please correct the typo.
  2. [Table II] The footnotes 'c' and 'd' state that the PointNet/PointNet++ [4] and SpikePoint accuracies use different class counts or data partitioning. Please move these qualifications into the main text or the table caption so that the comparison is not misread as being on identical test conditions.
  3. [V, Table III] The hyperparameters in Table III were 'selected to achieve an accuracy near 100% when tested on the training data.' This criterion risks overfitting to the training set; please report validation-set-based selection or at least discuss the sensitivity of the reported test accuracies to the threshold λ and the number of time steps K.
  4. [IV.A] The description of NMNIST as 'each consisting of 300 time samples' is imprecise; NMNIST samples have varying event counts and time durations. Please rephrase to match the dataset documentation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: accuracy is externally benchmarked and the energy numbers are arithmetic from stated formulas, with only a non-load-bearing inherited sparsity constant.

full rationale

PointLCA-Net's accuracy is evaluated on held-out test splits of NMNIST, DVS128, and SHD, so the central accuracy claims do not reduce to a fitted parameter or to the dictionary construction by definition. The energy estimate follows from explicit FLOP formulas (Eqs. 9-10) and an external RRAM energy constant from Yao et al. [37]. The only quantity inherited from the author's prior work is the 0.4% active-neuron fraction (Section IV.E: 'According to [28], up to 0.4% of neurons spike at each time step, and we have verified that this estimate is accurate'), which is a measured sparsity figure rather than a forced theorem, and the paper states it was independently checked. The self-citations [28] and [29] supply the Exemplar LCA-Decoder architecture and FLOP-counting framework, but they are not invoked as uniqueness results, and the current pipeline adds an external benchmark evaluation. The most serious concern is an energy-accounting incompleteness, not circularity: Section IV.E explicitly says the FLOP estimates 'exclude the FLOPs required for feature extraction by PointNet and PointNet++', while Table II compares the resulting 0.065 mJ against SpikePoint's full reported dynamic and static energy. If PointNet feature extraction were a large fraction of total cost, the 92% savings claim would be overstated; however, that is an omitted-cost/correctness issue, not a case of a prediction being equivalent to its input by construction. No specific circular step can be exhibited from the paper's equations.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central energy and accuracy claims rest on a chain of design choices: event-to-point-cloud conversion, untrained PointNet features, dictionary size equal to the training set, hand-chosen hyperparameters, and FLOPs-to-energy conversion. No new physical entities are introduced. The largest unpaid cost is PointNet feature extraction, which is excluded from the energy estimates.

free parameters (7)
  • Threshold lambda = 0.2
    Hand-chosen in Table III; Section V says hyperparameters were selected to achieve near 100% training accuracy.
  • Leakage tau = 1000
    Hand-selected in Table III; no sensitivity analysis is provided.
  • Number of time steps K = 100 (and 10 in Table I)
    Chosen by hand; affects inference FLOPs and accuracy.
  • Sliding window size and overlap = 0.5 s window, 0.25 s overlap
    DVS128 preprocessing choice in Section IV.D; no justification or sensitivity analysis.
  • Number of event points per sample = 1024
    Down-sampling target in Section IV.D; affects PointNet input size and accuracy.
  • Active neuron fraction = 0.4% average, with M-specific values 240, 114, 33
    Taken from reference [28] and asserted verified in Section IV.E; directly drives inference FLOPs and energy estimates.
  • Dictionary size M = Number of training samples: 60000, 28606, 8156
    The paper uses the entire training set as the dictionary (Section VI); this design choice drives both training and inference FLOPs.
assumptions (6)
  • domain assumption Spatio-temporal events can be represented as 3D point clouds with coordinates (x, y, t)
    Used throughout Sections III and IV.D; inherited from Wang et al. [4].
  • domain assumption Pretrained PointNet/PointNet++ features transfer to NMNIST, DVS128, and SHD without fine-tuning
    Section III.A assumes the global 1024-D features are discriminative for event data; the reported accuracies are the only evidence.
  • domain assumption The Exemplar LCA encoder-decoder with a dictionary of training features yields useful sparse codes
    Section III.B relies on [28] and [29], including the 0.4% active-neuron sparsity estimate.
  • standard math LCA neuron dynamics (Eq. 2 with threshold Eq. 6) converge to a sparse approximation
    Standard LCA result from Rozell et al. [30],[31], assumed without proof in Section III.B.
  • domain assumption FLOP counts in Eqs. 9-10 convert linearly to energy at 9.09e-14 J/FLOP on RRAM crossbars
    Section IV.E-F takes Yao et al. [37] RRAM efficiency and applies it to LCA operations without measured hardware.
  • ad hoc to paper PointNet feature extraction can be excluded from energy comparisons because it is performed once and stored
    Section IV.E explicitly excludes these FLOPs, but an end-to-end deployment still pays this cost, so the assumption materially affects the headline energy reduction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Spatio-Temporal Signal Recognition on Edge Devices Using PointLCA-Net." pith.science (2026). https://pith.science/paper/XDC6623V

@misc{pith2026241114585,
  author       = {Pith},
  title        = {Pith review of: Efficient Spatio-Temporal Signal Recognition on Edge Devices Using PointLCA-Net},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XDC6623V}},
  note         = {Machine review of arXiv:2411.14585}
}
read the original abstract

Recent advancements in machine learning, particularly through deep learning architectures like PointNet, have transformed the processing of three-dimensional (3D) point clouds, significantly improving 3D object classification and segmentation tasks. While 3D point clouds provide detailed spatial information, spatio-temporal signals introduce a dynamic element that accounts for changes over time. However, applying deep learning techniques to spatio-temporal signals and deploying them on edge devices presents challenges, including real-time processing, memory capacity, and power consumption. To address these issues, this paper presents a novel approach that combines PointNet's feature extraction with the in-memory computing capabilities and energy efficiency of neuromorphic systems for spatio-temporal signal recognition. The proposed method consists of a two-stage process: in the first stage, PointNet extracts features from the spatio-temporal signals, which are then stored in non-volatile memristor crossbar arrays. In the second stage, these features are processed by a single-layer spiking neural encoder-decoder that employs the Locally Competitive Algorithm (LCA) for efficient encoding and classification. This work integrates the strengths of both PointNet and LCA, enhancing computational efficiency and energy performance on edge devices. PointLCA-Net achieves high recognition accuracy for spatio-temporal data with substantially lower energy burden during both inference and training than comparable approaches, thus advancing the deployment of advanced neural architectures in energy-constrained environments.

Figures

Figures reproduced from arXiv: 2411.14585 by the authors.

Figure 1
Figure 1. PointLCA-Net Architecture: Features (ϕi) are extracted from the training data and stored in the synaptic weights of the Exemplar LCA-Decoder. The orange arrows represent the inference process, which follows the completion of training (feature extraction). varying signals. Neuromorphic systems can efficiently execute Vector￾Matrix Multiplication (VMM) operations, also known as Multiply-and-Accumulate (MAC) operations… view at source ↗
Figure 2
Figure 2. Feature Extraction using PointNets [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Data Pre-processing A. NMNIST The NMNIST [32] dataset is a spiking variant of the MNIST dataset, generated by presenting images to a neuromorphic vision camera equipped with the ATIS (Asynchronous Time￾based Image Sensor). This dataset comprises 60,000 training images, each consisting of 300 time samples, along with 10,000 test images used for accuracy evaluation. B. Spiking Heidelberg Digits (SHD) The SHD [33] data… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: PointLCA-Net Hardware Deployment Table I presents the estimated training and inference FLOPs for PointLCA-Net. “TFLOPs” refers to the Tera FLOPs re￾quired for training, while “GFLOPs” indicates the Giga FLOPs for inference operations. According to [28], up to 0.4% of n…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 40 canonical work pages

  1. [28]

    D-SELD: Dataset- Scalable Exemplar LCA-Decoder

    Mahmoodi Takaghaj, Sanaz, and Jack Sampson.“D-SELD: Dataset- Scalable Exemplar LCA-Decoder.” Neuromorphic Computing and En- gineering, 2024

    work page 2024

  2. [37]

    Joshua, and Qian, He

    Yao, Peng, Wu, Huaqiang, Gao, Bin, Tang, Jianshi, Zhang, Qingtian, Zhang, Wenqiang, Yang, J. Joshua, and Qian, He. Fully hardware- implemented memristor convolutional neural network . Nature, vol. 577, no. 7792, pp. 641–646, 2020. Nature Publishing Group UK London

    work page 2020

  3. [1]

    PointNet: Deep learning on point sets for 3D classification and segmentation,

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “PointNet: Deep learning on point sets for 3D classification and segmentation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pp. 652–660, 2017

    work page 2017

  4. [2]

    PointNet++: Deep hierarchical feature learning on point sets in a metric space,

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, “PointNet++: Deep hierarchical feature learning on point sets in a metric space,” Advances in Neural Information Processing Systems , vol. 30, 2017

    work page 2017

  5. [3]

    X. Ma, C. Qin, H. You, H. Ran, and Y . Fu, ”Rethinking network design and local geometry in point cloud: A simple residual MLP framework,” arXiv preprint arXiv:2202.07123 , 2022

    arXiv 2022

  6. [4]

    Space-time event clouds for gesture recognition: From RGB cameras to event cameras,

    Q. Wang, Y . Zhang, J. Yuan, and Y . Lu, “Space-time event clouds for gesture recognition: From RGB cameras to event cameras,” in 2019 IEEE Winter Conference on Applications of Computer Vision (WACV) , 2019, pp. 1826–1835

    work page 2019

  7. [5]

    S. Dave, R. Baghdadi, T. Nowatzki, S. Avancha, A. Shrivastava, and B. Li, ”Hardware acceleration of sparse and irregular tensor computations of ML models: A survey and insights,” Proc. IEEE, vol. 109, no. 10, pp. 1706–1752, 2021

    work page 2021

  8. [6]

    A., et al

    Merolla, P. A., et al. (2014). A million spiking-neuron integrated circuit with a scalable communication network and interface. Science, 345(6197), 668–673

    work page 2014

Show all 42 references
  1. [7]

    A. Amir, B. Taba, D. Berg, T. Melano, J. McKinstry, C. Di Nolfo, T. Nayak, A. Andreopoulos, G. Garreau, M. Mendoza, and others, ”A low power, fully event-based gesture recognition system,” in Proc. IEEE Conf. Comput. Vision Pattern Recognit. , 2017, pp. 7243–7252

    2017

  2. [8]

    Davies, N

    M. Davies, N. Srinivasa, T.-H. Lin, G. Chinya, Y . Cao, S. H. Choday, G. Dimou, P. Joshi, N. Imam, S. Jain, and others, ”Loihi: A neuromorphic manycore processor with on-chip learning,” IEEE Micro, vol. 38, no. 1, pp. 82–99, 2018

    2018

  3. [9]

    Davies, M., Wild, A., Orchard, G., Sandamirskaya, Y ., Guerra, G. A. F., Joshi, P., Plank, P., & Risbud, S. R. (2021). Advancing neuromorphic computing with Loihi: A survey of results and outlook. Proceedings of the IEEE, 109(5), 911–934. IEEE

    2021

  4. [10]

    S. B. Furber, F. Galluppi, S. Temple, and L. A. Plana, ”The SpiNNaker project,” Proc. IEEE, vol. 102, no. 5, pp. 652–665, 2014

    2014

  5. [11]

    Furber, S. (2016). Large-scale neuromorphic computing systems. Jour- nal of Neural Engineering , 13(5), 051001. IOP Publishing

    2016

  6. [12]

    The SpiNNaker 2 processing element archi- tecture for hybrid digital neuromorphic computing,

    S. H ¨oppner et al. , “The SpiNNaker 2 processing element archi- tecture for hybrid digital neuromorphic computing,” arXiv preprint arXiv:2103.08392, 2021

    2021 arXiv

  7. [13]

    Schmitt, S., Kl ¨ahn, J., Bellec, G., Gr ¨ubl, A., Guettler, M., Hartel, A., Hartmann, S., Husmann, D., Husmann, K., Jeltsch, S., et al. (2017). Neuromorphic hardware in the loop: Training a deep spiking network on the BrainScaleS wafer-scale system. In 2017 International Join...

    2017

  8. [14]

    Pehle, C., Billaudelle, S., Cramer, B., Kaiser, J., Schreiber, K., Strad- mann, Y ., Weis, J., Leibfried, A., M¨uller, E., & Schemmel, J. (2022). The BrainScaleS-2 accelerated neuromorphic system with hybrid plasticity. Frontiers in Neuroscience, 16, 795876. Frontiers Media SA

    2022

  9. [15]

    J. Pei, L. Deng, S. Song, M. Zhao, Y . Zhang, S. Wu, G. Wang, Z. Zou, Z. Wu, W. He, and others, ”Towards artificial general intelligence with hybrid Tianjic chip architecture,” Nature, vol. 572, no. 7767, pp. 106–111, 2019

    2019

  10. [16]

    M., K ¨ostinger, G., Nielsen, C., Qiao, N., & Indiveri, G

    Richter, O., Wu, C., Whatley, A. M., K ¨ostinger, G., Nielsen, C., Qiao, N., & Indiveri, G. (2024). DYNAP-SE2: A scalable multi-core dynamic neuromorphic asynchronous spiking neural network processor. Neuro- morphic Computing and Engineering , 4(1), 014003. IOP Publishing

    2024

  11. [17]

    Ghosh-Dastidar, S., & Adeli, H. (2009). Spiking neural networks. Inter- national Journal of Neural Systems , 19(04), 295–308. World Scientific

    2009

  12. [18]

    Gr ¨uning, A., & Bohte, S. M. (2014). Spiking neural networks: Principles and challenges. In ESANN

    2014

  13. [19]

    NVM neuromorphic core with 64k-cell (256-by-256) phase change memory synaptic array with on-chip neuron circuits for continuous in-situ learning,

    S. Kim et al. , “NVM neuromorphic core with 64k-cell (256-by-256) phase change memory synaptic array with on-chip neuron circuits for continuous in-situ learning,” in Proc. Int. Electron Devices Meeting (IEDM), 2015, pp. 17–1

    2015

  14. [20]

    ”Nanoelectronic Programmable Synapses Based on Phase Change Materials for Brain-Inspired Computing.” Nano Letters, vol

    Kuzum, Duygu, Rakesh GD Jeyasingh, Byoungil Lee, and H-S Philip Wong. ”Nanoelectronic Programmable Synapses Based on Phase Change Materials for Brain-Inspired Computing.” Nano Letters, vol. 12, no. 5, 2012, pp. 2179-2186. ACS Publications

    2012

  15. [21]

    Detecting correlations using phase-change neurons and synapses,

    T. Tuma, M. Le Gallo, A. Sebastian, and E. Eleftheriou, “Detecting correlations using phase-change neurons and synapses,” IEEE Electron Device Lett., vol. 37, no. 10, pp. 1238–1241, Oct. 2016

    2016

  16. [22]

    Analog memristive synapse in spiking networks imple- menting unsupervised learning,

    E. Covi et al., “Analog memristive synapse in spiking networks imple- menting unsupervised learning,” Front. Neurosci., vol. 10, p. 482, 2016

    2016

  17. [23]

    All-memristive neuromorphic computing with level-tuned neurons,

    A. Pantazi, S. Wo ´zniak, T. Tuma, and E. Eleftheriou, “All-memristive neuromorphic computing with level-tuned neurons,” Nanotechnology, vol. 27, no. 35, p. 355205, 2016, IOP Publishing

    2016

  18. [24]

    M. Rao, H. Tang, J. Wu, W. Song, M. Zhang, W. Yin, Y . Zhuo, F. Kiani, B. Chen, X. Jiang, and others, ”Thousands of conductance levels in memristors integrated on CMOS,” Nature, vol. 615, no. 7954, pp. 823–829, 2023

    2023

  19. [25]

    Unsupervised learning in probabilistic neural networks with multi-state metal-oxide memristive synapses,

    A. Serb et al., “Unsupervised learning in probabilistic neural networks with multi-state metal-oxide memristive synapses,” Nature Communica- tions, vol. 7, p. 12611, 2016

    2016

  20. [26]

    Memory devices and applications for in- memory computing

    Sebastian, Abu, Le Gallo, Manuel, Khaddam-Aljameh, Riduan, and Eleftheriou, Evangelos. “Memory devices and applications for in- memory computing.” Nature Nanotechnology 15, no. 7 (2020): 529–544. Nature Publishing Group UK London

    2020

  21. [27]

    Memory and information pro- cessing in neuromorphic systems

    Indiveri, Giacomo, and Liu, Shih-Chii. “Memory and information pro- cessing in neuromorphic systems.” Proceedings of the IEEE 103, no. 8 (2015): 1379–1397. IEEE

    2015

  22. [29]

    ViT-LCA: A Neuromorphic Approach for Vision Transformers

    Mahmoodi Takaghaj, Sanaz. “ViT-LCA: A Neuromorphic Approach for Vision Transformers.” arXiv preprint arXiv:2411.00140 , 2024

    2024 arXiv

  23. [30]

    Sparse coding via thresholding and local competition in neural circuits,

    C. J. Rozell, D. H. Johnson, R. G. Baraniuk, and B. A. Olshausen, “Sparse coding via thresholding and local competition in neural circuits,” Neural Computation, vol. 20, no. 10, pp. 2526–2563, 2008

    2008

  24. [31]

    Locally Com- petitive Algorithms for Sparse Approximation,

    C. Rozell, D. Johnson, R. Baraniuk, and B. Olshausen, “Locally Com- petitive Algorithms for Sparse Approximation,” in Proc. 2007 IEEE Int. Conf. Image Processing , vol. 4, pp. IV-169–IV-172, 2007, doi: 10.1109/ICIP.2007.4379981

    2007

  25. [32]

    Converting Static Image Datasets to Spiking Neuromorphic Datasets Using Sac- cades,

    G. Orchard, A. Jayawant, G. K. Cohen, and N. Thakor, “Converting Static Image Datasets to Spiking Neuromorphic Datasets Using Sac- cades,” Frontiers in Neuroscience, vol. 9, p. 437, 2015.“

    2015

  26. [33]

    The Heidelberg Spiking Data Sets for the Systematic Evaluation of Spiking Neural Net- works,

    B. Cramer, Y . Stradmann, J. Schemmel, and F. Zenke, “The Heidelberg Spiking Data Sets for the Systematic Evaluation of Spiking Neural Net- works,” IEEE Transactions on Neural Networks and Learning Systems , vol. 33, no. 7, pp. 2744–2757, 2020

    2020

  27. [34]

    Spiking pointnet: Spiking neural networks for point clouds,

    D. Ren, Z. Ma, Y . Chen, W. Peng, X. Liu, Y . Zhang, and Y . Guo, “Spiking pointnet: Spiking neural networks for point clouds,” Advances in Neural Information Processing Systems , vol. 36, 2024

    2024

  28. [35]

    SpikePoint: An Efficient Point-based Spiking Neural Network for Event Cameras Action Recognition,

    H. Ren, Y . Zhou, X. Lin, Y . Huang, H. Fu, J. Song, and B. Cheng, “SpikePoint: An Efficient Point-based Spiking Neural Network for Event Cameras Action Recognition,” in Proc. of the Twelfth International Conference on Learning Representations , 2024

    2024

  29. [36]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, and others, ”PyTorch: An imperative style, high-performance deep learning library,” Advances in Neural Information Processing Systems , vol. 32, 2019

    2019

  30. [38]

    Point Trans- former,

    H. Zhao, L. Jiang, J. Jia, P. H. S. Torr, and V . Koltun, “Point Trans- former,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 16259–16268, 2021

    2021

  31. [39]

    ”Go Wider: An Efficient Neural Network for Point Cloud Analysis via Group Convolutions.” Applied Sciences, vol

    Chen, Can, Luca Zanotti Fragonara, and Antonios Tsourdos. ”Go Wider: An Efficient Neural Network for Point Cloud Analysis via Group Convolutions.” Applied Sciences, vol. 10, no. 7, 2020, p. 2391. MDPI

    2020

  32. [40]

    Khaddam-Aljameh, M

    R. Khaddam-Aljameh, M. Stanisavljevic, J. F. Mas, G. Karunaratne, M. Br¨andli, F. Liu, A. Singh, S. M. M ¨uller, U. Egger, A. Petropoulos, and others, ”HERMES-core—A 1.59-TOPS/mm² PCM on 14-nm CMOS in- memory compute core using 300-ps/LSB linearized CCO-based ADCs,” IEEE J. So...

    2022

  33. [41]

    Le Gallo, R

    M. Le Gallo, R. Khaddam-Aljameh, M. Stanisavljevic, A. Vasilopoulos, B. Kersting, M. Dazzi, G. Karunaratne, M. Br ¨andli, A. Singh, S. M. Mueller, and others, ”A 64-core mixed-signal in-memory compute chip based on phase-change memory for deep neural network inference,” Nature...

    2023

  34. [42]

    Joshua, Lu, Wei, Chang, Meng-Fan, Ielmini, Daniele, Yang, Yuchao, et al

    Aguirre, Fernando, Sebastian, Abu, Le Gallo, Manuel, Song, Wenhao, Wang, Tong, Yang, J. Joshua, Lu, Wei, Chang, Meng-Fan, Ielmini, Daniele, Yang, Yuchao, et al. Hardware implementation of memristor- based artificial neural networks . Nature Communications, vol. 15, no. 1, p. 1...

    1974

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.