Pith. sign in

REVIEW 4 major objections 7 minor 38 references

Histogram-Equalized Quantization for logic-gated Residual Neural Networks

T0 review · 4 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read A simple quantile-based step size rule can match learned quantization and enable logic-gated residual networks.

desk verdict Neat quantile-based step-size rule and clever logic-gated residuals, but the empirical claims rest on a 0.01% margin and unmatched baselines. read the letter →

arxiv 2501.04517 v2 pith:F3LOXMFH submitted 2025-01-08 cs.LG cs.AR

classification cs.LGcs.AR
keywords quantizedneuralnetworkshistogramequalizationquantization-awaretraininglinearsymmetricquantizationternaryweightslogic-gatedresidualskipconnectionshardware-efficientinference
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 Histogram-Equalized Quantization (HEQ), a way to choose the step size of linear symmetric quantized neural networks by looking at the layer-wise histogram of full-precision proxy weights. Instead of learning the step size through backpropagation, HEQ finds the n-quantiles of the weight distribution and sets the step size so that the quantized values fall close to those quantiles, balancing how often each discrete value is used. The authors report that this rule matches or exceeds learned-step-size methods on CIFAR-10, reaching 93.51% accuracy with 2-bit weights and activations, and no loss compared to the full-precision baseline when using 5 or 7 weight levels. On STL-10, the same rule makes it possible to replace full-precision residual additions with OR and MUX logic gates, giving 84.17% accuracy with a smaller model than a budget-aware learned-step-size baseline. The practical payoff would be cheaper hardware: quantized weights are exactly 0, ±1, or ±0.5, so multiplications become logic operations and bit shifts.

What carries the argument

The central object is the quantile-matched step size. At the start of each epoch, each layer's full-precision proxy weights are sorted into n-quantiles $q_{-i}, q_i$; assuming $q_{-i} \approx -q_i$, the step size is set by $$s = \frac{4 \sum_{i=1}^{(n-1)/2}(|q_{-i}|+q_i)}{(n-1)^2},$$ which is exactly the value that makes the quantization thresholds $(2i-1)s/2$ coincide with those quantiles. This turns the histogram into the quantizer's calibration signal, replacing learned gradients or fixed norm factors, and it is the mechanism that makes the quantized weights land on 0, ±1, and ±0.5, values that map directly to logic gates and bit shifts in hardware.

What would settle it

Train the same CIFAR-10 model with weights initialized from a distribution with a nonzero per-layer mean, measure the asymmetry $\sum_i(|q_{-i}| - q_i)$ at every epoch, and check both the quantized-bin balance and final accuracy; if layers with strong asymmetry show unbalanced bins and degraded accuracy despite HEQ, the symmetry assumption is the bottleneck.

Watch

Extended reading notes

Core claim

The central claim is that a proper quantizer should spread the quantized weights evenly across the available discrete levels, and that this equalization can be achieved without learned parameters. For each layer, HEQ computes the n-quantiles of the current proxy weight histogram, assumes the positive and negative quantiles are mirror images around zero, and updates the single step size s with Eq. (3) so that the quantization thresholds coincide with those quantiles. Because the update is recomputed once per epoch from the histogram, the step size tracks the evolving distribution during training. Empirically this yields balanced ternary and quinary weight histograms, CIFAR-10 accuracy of 93.51% for 2/2-bit W/A quantization and 93.75% for septenary weights, and, on STL-10, proper training of residual networks whose skip connections are implemented by OR and MUX gates rather than 32-bit additions, reaching 84.17% accuracy.

Load-bearing premise

The load-bearing premise is that each layer's weight distribution is symmetric enough around zero that the positive and negative quantile pairs cancel in the step-size formula; if the histogram is skewed, the step size is miscalibrated and the intended equalization is not achieved.

Editorial extensions

If this is right

  • On CIFAR-10 with VGG-Small, HEQ-ternary reaches 93.51% accuracy at 2-bit weights and 2-bit activations, on par with learned quantization methods while using no learned step size.
  • HEQ-quinary and HEQ-septenary reach 93.66% and 93.75% respectively, matching or slightly exceeding the 93.68% full-precision baseline.
  • On STL-10, HEQ-ternary with the OR-gated variant reaches 83.82% and with the MUX-OR variant 84.17%, beating a regularized learned-step-size baseline that uses more parameters.
  • The quantized weights are exactly 0, ±1, or ±0.5, so hardware can implement them with logic operations and bit shifts rather than full-precision multipliers and scale factors.
  • Replacing residual additions and attention-like multiplications by OR and MUX gates is compatible with HEQ, reducing the hardware cost of skip connections without sacrificing accuracy.

Reading between the lines

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

  • The rule is stated only for odd n and for weights; extending the same quantile matching to activations, or to even n, is a natural test that the paper does not run.
  • If the symmetry assumption fails, a simple fix would be to recenter each layer's weights before computing quantiles; that would let HEQ work on skewed distributions and is directly measurable.
  • The logic-gate result suggests that full-precision skip connections are not what makes residual training stable once the weight histogram is well calibrated; this could change how residual blocks are mapped to hardware.
  • One testable consequence is that HEQ's advantage over learned step sizes should grow with the bitwidth, since quantile matching becomes more informative as n increases.
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 / 7 minor

Summary. The paper proposes Histogram-Equalized Quantization (HEQ), an adaptive linear symmetric quantization method that updates the per-layer step size at the beginning of each epoch using n-quantiles of the proxy weight distribution, with the goal of equalizing the histogram of quantized values. The method is applied to ternary, quinary, and septenary weights with 2-bit activations on CIFAR-10, where the authors report 93.51% accuracy for HEQ-ternary and claim state-of-the-art performance. The paper also introduces OR-gated and MUX-gated residual blocks intended to replace full-precision skip connections, and reports STL-10 accuracies of 83.34% (VGG-11), 83.82% (ORNet-11), and 84.17% (MUXORNet-11). The central claims are that HEQ automatically adapts quantization thresholds without learned step sizes and that it enables effective training of logic-gated residual networks at lower hardware complexity.

Significance. If the empirical claims hold, HEQ would be an attractively simple replacement for learned step-size quantization, requiring no extra parameters and producing integer-only weights (0, ±1, ±0.5) that are directly hardware-friendly. The logic-gated residual network idea is also interesting because it could eliminate full-precision additions in ResNet-style topologies. The paper is clearly written and the algorithm is easy to reproduce. However, the evidence presented is not yet sufficient: the CIFAR-10 state-of-the-art claim rests on a 0.01 percentage point margin with no error bars, and the STL-10 comparison varies architecture, parameter count, bitwidth, and training procedure simultaneously. The symmetry assumption underlying Eq. (3) is not validated. The significance is therefore conditional on stronger empirical support and a more controlled experimental design.

major comments (4)
  1. [Section III-A, Eq. (3)] The derivation of the step-size update is built on the assumption q_{-i} ≈ -q_i, i.e., that layer-wise weight quantiles are symmetric around zero. This assumption is stated but not validated. If a layer's weight distribution is skewed, the computed s will not equalize the quantization bins, and the method's core rationale is undermined. Please provide quantitative evidence of symmetry across layers and training stages (e.g., skewness of the proxy weight histograms, or a comparison of q_{-i} and -q_i), or report the performance degradation on a deliberately skewed distribution.
  2. [Table I, CIFAR-10] The claim that HEQ achieves state-of-the-art accuracy on CIFAR-10 is based on HEQ-ternary at 93.51% versus LQ at 93.50% for 2/2 bit width, a difference of 0.01 percentage points. The paper reports averages over 5 realizations but gives no standard deviations, confidence intervals, or significance tests. This margin is well within typical run-to-run variance for CIFAR-10 training, so the result does not demonstrate superiority. Please report the per-realization results with error bars and, if possible, a paired significance test, or soften the state-of-the-art claim to a claim of competitiveness.
  3. [Table II, STL-10] The STL-10 comparison changes multiple factors at once: the reference method [18] uses a VGG-7 backbone (4.57M parameters) with 2.5/8 bit width, while the proposed models use a VGG-11 backbone (3.14M parameters) with 2/1 bit width. Architecture, parameter count, bit width, and training procedure all differ, so the 0.37% improvement of MUXORNet-11 over [18] cannot be attributed to HEQ or to the OR/MUX gates. A controlled comparison is needed, for example VGG-7 with HEQ versus [18] under matching bit widths, or VGG-11 with and without logic gates under otherwise identical settings.
  4. [Section IV-B and Fig. 5] There is no ablation that isolates the contribution of HEQ from the contribution of the OR/MUX gating. In particular, the paper does not compare ORNet-11 or MUXORNet-11 against the same VGG-11 backbone with a standard full-precision residual addition, nor against a version trained with a learned step-size quantizer. Without such ablations, the claim that 'HEQ enables a proper training' of logic-gated residual networks is not supported. Please add these baselines.
minor comments (7)
  1. [Fig. 1] The first two subfigure captions both read 'Full-precision model.' Please distinguish them, for instance by indicating the layer index or the weight/activation type.
  2. [Eq. (1)] The notation for the rounding/flooring operation in Eq. (1) is garbled in the text. Please define it explicitly, e.g., as \lfloor w/s \rfloor or nearest-integer rounding.
  3. [Algorithm 1] The instruction 'Find n-quantiles of layer l' is underspecified. Please state the quantile estimator used and whether the quantiles are computed on the proxy weights or on the quantized weights.
  4. [Section III-A] The sentence 'this can be enhanced by forcing the weight median to zero, which has not been applied' is a self-acknowledged limitation. Please either implement the median-centering or provide evidence that it has negligible impact.
  5. [Section IV-A] The MUX threshold uses a fixed factor of 0.5. This is an additional free parameter; please report a sensitivity analysis or at least justify the choice.
  6. [Section III-B] The paper motivates HEQ by an entropy-maximization argument but reports no entropy statistics. Reporting the entropy of quantized weights for HEQ versus baselines would substantiate the mechanism.
  7. [Table I] Please clarify why HEQ-septenary is marked '+' rather than '++' in the HW-Compatibility column, and why a 3-bit representation is used for 7 levels rather than 8.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HEQ's step-size rule is a construction, and the accuracy claims are external empirical benchmarks.

full rationale

I found no circular step in the paper. HEQ's step-size update in Eq. (3) is constructed from the layer-wise weight quantiles so that the quantization thresholds approximately coincide with equal-probability intervals, under the stated symmetry assumption q_-i ≈ -q_i. This is a design rule rather than a prediction: the balanced histograms shown in Fig. 1 are consequences of the construction, and the accuracy numbers in Tables I and II are measured against external benchmarks and competing published methods. No load-bearing claim reduces to a fitted parameter renamed as a prediction: s is updated from the weights' own quantiles, not from the accuracy target, and the reported accuracies are independent external metrics. There are no self-citations used as evidence; the cited prior works are independent of the authors, and no uniqueness theorem or ansatz is imported from the authors' own earlier work. The empirical limitations noted by the skeptic—the 0.01% CIFAR-10 margin without significance testing and the multi-factor confounding in the STL-10 comparison—are evidentiary or correctness concerns, not circularity. Because the derivation chain is self-contained and the central claims are evaluated empirically, the circularity score is 0.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

HEQ itself introduces no fitted scalar constants; the step size s is computed directly from weight quantiles. The MUX threshold factor 0.5 is a hand-set constant in the logic-gated branch. The key assumptions are the entropy hypothesis and the symmetry of weight quantiles, along with standard QAT practices such as STE.

free parameters (1)
  • MUX threshold factor 0.5 = 0.5
    In the MUX branch, the rebinarization threshold is T(x)=1{x>0.5m} with m=1 in quantized models; this constant is hand-chosen, not fitted to accuracy, and affects the OR/MUX balancing behavior.
assumptions (4)
  • domain assumption A proper quantizer should maximize the entropy of quantized weights, i.e., cover the full representation space.
    Sections III intro and III-A motivate HEQ from this hypothesis; it is not proven and is the basis of the equalization objective.
  • domain assumption Proxy weight quantiles are symmetrically distributed around zero (q_{-i} ≈ -q_i).
    Used in Eq. (2)-(3); if violated, the step size update is biased and the intended equalization is not achieved.
  • domain assumption The Straight-Through Estimator provides a usable gradient for the quantization step.
    Section II uses STE for backpropagation; this is a standard approximation in QAT but is not exact.
  • domain assumption Quantiles computed once per epoch are stable predictors of the next epoch's weight distribution.
    Algorithm 1 updates s only at epoch boundaries; if distributions shift rapidly, the step size lags behind the current weights.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Histogram-Equalized Quantization for logic-gated Residual Neural Networks." pith.science (2026). https://pith.science/paper/F3LOXMFH

@misc{pith2026250104517,
  author       = {Pith},
  title        = {Pith review of: Histogram-Equalized Quantization for logic-gated Residual Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F3LOXMFH}},
  note         = {Machine review of arXiv:2501.04517}
}
read the original abstract

Adjusting the quantization according to the data or to the model loss seems mandatory to enable a high accuracy in the context of quantized neural networks. This work presents Histogram-Equalized Quantization (HEQ), an adaptive framework for linear symmetric quantization. HEQ automatically adapts the quantization thresholds using a unique step size optimization. We empirically show that HEQ achieves state-of-the-art performances on CIFAR-10. Experiments on the STL-10 dataset even show that HEQ enables a proper training of our proposed logic-gated (OR, MUX) residual networks with a higher accuracy at a lower hardware complexity than previous work.

Figures

Figures reproduced from arXiv: 2501.04517 by the authors.

Figure 1
Figure 1. Weight distributions of 2 layers (in 2 columns) after [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 4
Figure 4. depicts the variation of s during training in both ternary and quinary cases. It shows that the evolution of s depends on the layer and has different convergence values. (a) Ternary quantization. (b) Quinary quantization [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figure 3
Figure 3. Comparison of the ternary-weight distribution using [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Models with the plain block (11-hidden layer VGG [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 21 canonical work pages

  1. [18]

    Adaptive Quantization Method for CNN with Computational-Complexity-Aware Regularization,

    K. Nakata, D. Miyashita, J. Deguchi, and R. Fujimoto, “Adaptive Quantization Method for CNN with Computational-Complexity-Aware Regularization,” in 2021 IEEE International Symposium on Circuits and Systems (ISCAS), May 2021, pp. 1–5, iSSN: 2158-1525

  2. [1]

    Quantized neural networks: Training neural networks with low pre- cision weights and activations,

    I. Hubara, M. Courbariaux, D. Soudry, R. El-Yaniv, and Y . Bengio, “Quantized neural networks: Training neural networks with low pre- cision weights and activations,” J. Mach. Learn. Res. , vol. 18, no. 1, 2017

  3. [2]

    UNPU: An Energy-Efficient Deep Neural Network Accelerator With Fully Variable Weight Bit Precision,

    J. Lee, C. Kim, S. Kang, D. Shin, S. Kim, and H.-J. Yoo, “UNPU: An Energy-Efficient Deep Neural Network Accelerator With Fully Variable Weight Bit Precision,” IEEE Journal of Solid-State Circuits , vol. 54, no. 1, pp. 173–185, Jan. 2019

  4. [3]

    A 617 TOPS/W All Digital Binary Neural Network Accelerator in 10nm FinFET CMOS,

    P. C. Knag, G. K. Chen, H. E. Sumbul, R. Kumar, M. A. Anders, H. Kaul, S. K. Hsu, A. Agarwal, M. Kar, S. Kim, and R. K. Krishna- murthy, “A 617 TOPS/W All Digital Binary Neural Network Accelerator in 10nm FinFET CMOS,” in 2020 IEEE Symposium on VLSI Circuits , Jun. 2020, pp. 1–2, iSSN: 2158-5636

  5. [4]

    A Resource-Efficient Inference Accelerator for Binary Convolutional Neural Networks,

    T.-H. Kim and J. Shin, “A Resource-Efficient Inference Accelerator for Binary Convolutional Neural Networks,” IEEE Transactions on Circuits and Systems II: Express Briefs , vol. 68, no. 1, pp. 451–455, Jan. 2021, conference Name: IEEE Transactions on Circuits and Systems II: Express Briefs

  6. [5]

    Chewbaccann: A flexible 223 tops/w bnn accelerator,

    R. Andri, G. Karunaratne, L. Cavigelli, and L. Benini, “Chewbaccann: A flexible 223 tops/w bnn accelerator,” 2021 IEEE International Sym- posium on Circuits and Systems (ISCAS) , pp. 1–5, 2021

  7. [6]

    A convolutional neu- ral network accelerator architecture with fine-granular mixed precision configurability,

    X. Zhou, L. Zhang, C. Guo, X. Yin, and C. Zhuo, “A convolutional neu- ral network accelerator architecture with fine-granular mixed precision configurability,” in 2020 IEEE International Symposium on Circuits and Systems (ISCAS), 2020, pp. 1–5

  8. [7]

    Binarized neural networks: Training deep neural networks with weights and activations constrained to +1 or -1,

    M. Courbariaux, I. Hubara, D. Soudry, R. El-Yaniv, and Y . Bengio, “Binarized neural networks: Training deep neural networks with weights and activations constrained to +1 or -1,” arXiv: Learning, 2016

Show all 38 references
  1. [8]

    Deep compression: Compressing deep neural network with pruning, trained quantization and huffman coding,

    S. Han, H. Mao, and W. Dally, “Deep compression: Compressing deep neural network with pruning, trained quantization and huffman coding,” arXiv: Computer Vision and Pattern Recognition , 2016

  2. [9]

    Convolutional neu- ral networks using logarithmic data representation,

    D. Miyashita, E. H. Lee, and B. Murmann, “Convolutional neu- ral networks using logarithmic data representation,” ArXiv, vol. abs/1603.01025, 2016

  3. [10]

    Lq-nets: Learned quantization for highly accurate and compact deep neural networks,

    D. Zhang, J. Yang, D. Ye, and G. Hua, “Lq-nets: Learned quantization for highly accurate and compact deep neural networks,” ArXiv, vol. abs/1807.10029, 2018

  4. [11]

    Linear symmetric quantization of neural networks for low-precision integer hardware,

    X. Zhao, Y . Wang, X. Cai, C. Liu, and L. Zhang, “Linear symmetric quantization of neural networks for low-precision integer hardware,” in ICLR, 2020

  5. [12]

    Ternary weight networks,

    F. Li and B. Liu, “Ternary weight networks,” ArXiv, vol. abs/1605.04711, 2016

  6. [13]

    Trained ternary quantization,

    C. Zhu, S. Han, H. Mao, and W. J. Dally, “Trained ternary quantization,” in 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings. OpenReview.net, 2017

  7. [14]

    TRQ: Ternary Neural Networks With Residual Quantization,

    Y . Li, W. Ding, C. Liu, B. Zhang, and G. Guo, “TRQ: Ternary Neural Networks With Residual Quantization,” Proceedings of the AAAI Conference on Artificial Intelligence , vol. 35, no. 10, pp. 8538–8546, May 2021, number: 10

  8. [15]

    Learned step size quantization,

    S. K. Esser, J. L. McKinstry, D. Bablani, R. Appuswamy, and D. S. Modha, “Learned step size quantization,” in 8th International Confer- ence on Learning Representations, ICLR , 2020

  9. [16]

    Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation,

    Y . Bengio, N. L ´eonard, and A. Courville, “Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation,” arXiv:1308.3432 [cs], Aug. 2013

  10. [17]

    Mixed precision DNNs: All you need is a good parametrization,

    S. Uhlich, L. Mauch, F. Cardinaux, K. Yoshiyama, J. A. Garc ´ıa, S. Tiedemann, T. Kemp, and A. Nakamura, “Mixed precision DNNs: All you need is a good parametrization,” in 8th International Conference on Learning Representations, ICLR, Addis Ababa, Ethiopia, April 26-30, 2020

  11. [19]

    A mathematical theory of communication,

    C. E. Shannon, “A mathematical theory of communication,” The Bell System Technical Journal, vol. 27, pp. 379–423, 1948

  12. [20]

    Effective quantization methods for recurrent neural networks,

    Q. He, H. Wen, S. Zhou, Y . Wu, C. Yao, X. Zhou, and Y . Zou, “Effective quantization methods for recurrent neural networks,” ArXiv, vol. abs/1611.10176, 2016

  13. [21]

    Effective Quantization Approaches for Recurrent Neural Networks,

    M. Z. Alom, A. T. Moody, N. Maruyama, B. C. Van Essen, and T. M. Taha, “Effective Quantization Approaches for Recurrent Neural Networks,” in 2018 International Joint Conference on Neural Networks (IJCNN). Rio de Janeiro: IEEE, Jul. 2018, pp. 1–8. [Online]. Available: https://i...

  14. [22]

    Dorefa-net: Training low bitwidth convolutional neural networks with low bitwidth gradients,

    S. Zhou, Z. Ni, X. Zhou, H. Wen, Y . Wu, and Y . Zou, “Dorefa-net: Training low bitwidth convolutional neural networks with low bitwidth gradients,” ArXiv, vol. abs/1606.06160, 2016

  15. [23]

    Learning Multiple Layers of Features from Tiny Images,

    A. Krizhevsky, “Learning Multiple Layers of Features from Tiny Images,” p. 60, 2009. [Online]. Available: https://www.cs.toronto.edu/ ∼kriz/learning-features-2009-TR.pdf

  16. [24]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift,

    S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” ArXiv, vol. abs/1502.03167, 2015

  17. [25]

    Soft threshold ternary networks,

    W. Xu, X. He, T. Zhao, Q. Hu, P. Wang, and J. Cheng, “Soft threshold ternary networks,” in IJCAI, 2020

  18. [26]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2016

  19. [27]

    Bi-real net: Enhancing the performance of 1-bit cnns with improved representational capability and advanced training algorithm,

    Z. Liu, B. Wu, W. Luo, X. Yang, W. Liu, and K. Cheng, “Bi-real net: Enhancing the performance of 1-bit cnns with improved representational capability and advanced training algorithm,” in ECCV, 2018

  20. [28]

    Mobinet: A mobile binary network for image classification,

    H. T. Phan, D. Huynh, Y . He, M. Savvides, and Z. Shen, “Mobinet: A mobile binary network for image classification,” 2020 IEEE Winter Conference on Applications of Computer Vision (WACV) , pp. 3442– 3451, 2020

  21. [29]

    Hawqv3: Dyadic neural network quantization,

    Z. Yao, Z. Dong, Z. Zheng, A. Gholami, J. Yu, E. Tan, L. Wang, Q. Huang, Y . Wang, M. W. Mahoney, and K. Keutzer, “Hawqv3: Dyadic neural network quantization,” in ICML, 2021

  22. [30]

    Residual attention network for image classification,

    F. Wang, M. Jiang, C. Qian, S. Yang, C. Li, H. Zhang, X. Wang, and X. Tang, “Residual attention network for image classification,” 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 6450–6458, 2017

  23. [31]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” in 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings , Y . Bengio and Y . LeCun, Eds.,

  24. [32]

    Embedded floating-point units in FPGAs,

    M. Beauchamp, S. Hauck, K. Underwood, and K. Hemmert, “Embedded floating-point units in FPGAs,” in Proceedings of the 2006 ACM/SIGDA 14th international symposium on Field programmable gate array , 01 2006, pp. 12–20

  25. [33]

    Computing’s energy problem (and what we can do about it),

    M. Horowitz, “Computing’s energy problem (and what we can do about it),” 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC), pp. 10–14, 2014

  26. [34]

    An analysis of single-layer networks in unsupervised feature learning,

    A. Coates, A. Ng, and H. Lee, “An analysis of single-layer networks in unsupervised feature learning,” in AISTATS, 2011

  27. [35]

    Improved regularization of convolutional neural networks with cutout,

    T. Devries and G. W. Taylor, “Improved regularization of convolutional neural networks with cutout,” ArXiv, vol. abs/1708.04552, 2017

  28. [36]

    TensorFlow: Large-scale machine learning on heterogeneous systems,

    M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C. Citro, G. S. Corrado, A. Davis, J. Dean, M. Devin, S. Ghemawat, I. Goodfellow, A. Harp, G. Irving, M. Isard, Y . Jia, R. Jozefowicz, L. Kaiser, M. Kudlur, J. Levenberg, D. Man ´e, R. Monga, S. Moore, D. Murray, C. Olah, M...

  29. [37]

    Larq: An open-source library for training binarized neural networks,

    L. Geiger and P. Team, “Larq: An open-source library for training binarized neural networks,” J. Open Source Softw., vol. 5, p. 1746, 2020

  30. [2015]

    Available: http://arxiv.org/abs/1409.1556

    [Online]. Available: http://arxiv.org/abs/1409.1556

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.