Pith. sign in

REVIEW 3 major objections 5 minor 38 references

Differentiable Learning-to-Group Channels via Groupable Convolutional Neural Networks

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper shows that convolutional networks can learn per-layer channel-group counts, and that doing so beats fixed-group ResNeXt at matched or lower cost.

desk verdict Learned per-layer group counts transfer to standard ResNeXt and beat the constant 32×4d baseline; a missing random non-uniform control keeps the mechanism claim conditional. read the letter →

arxiv 1908.05867 v2 pith:IDBJH6SN submitted 2019-08-16 cs.CV cs.LGcs.NE

Zhaoyang Zhang , Jingyu Li , Wenqi Shao , Zhanglin Peng , Ruimao Zhang , Xiaogang Wang , Ping Luo This is my paper · ORCID
classification cs.CVcs.LGcs.NE
keywords groupconvolutiondynamicgroupinglearnedcardinalityResNeXtstraight-throughestimatorKroneckerproductresource-constrainedtrainingImageNet
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

Group convolution is a standard way to cut cost in ConvNets, but the number of groups is usually a hand-set constant that is the same in every layer. This paper argues that constant cardinality is sub-optimal and proposes Dynamic Grouping Convolution (DGConv), which learns per-layer grouping strategies end-to-end. The authors build Groupable ResNeXt by replacing every group convolution with DGConv, and report that it outperforms re-implemented ResNet and ResNeXt on ImageNet with equal or fewer parameters and fewer group-convolution FLOPs. The learned per-layer group counts also transfer to ordinary group convolution, so the final model needs no special inference code. If correct, the result is a new, cheaply optimizable architectural dimension: where and how to group channels.

What carries the argument

The central mechanism is a Kronecker-structured binary relationship matrix $U$ that masks the convolution kernel. For a layer with $C$ input and output channels, $U$ is built as $U = U_1 \otimes \cdots \otimes U_K$ with $K = \log_2 C$ factors, each $U_k = g_k \mathbf{1} + (1-g_k) I$, where $g_k = \operatorname{sign}(\tilde{g}_k)$ and $\tilde{g}_k$ is a continuous learnable gate. This construction uses only $\log_2 C$ parameters per layer, always yields a valid group structure (adjacent or non-adjacent channels), and covers regular, group, and depthwise convolution as special cases. The sign function is trained with a straight-through estimator, and the complexity penalty $\zeta = \sum_l C_l \prod_k (1 + g^l_k)$ lets the user cap the total number of active kernel connections. That combination is what makes the group count itself a differentiable, resource-controllable parameter.

What would settle it

One decisive test: train two Groupable ResNeXts of the same budget with all gates initialized positive versus all negative. If their learned per-layer group counts differ substantially while validation accuracy matches, the grouping strategy is determined more by optimization path than by the data, and the claimed gain is not attributable to the learned structure. A simpler check is to replace the learned group counts in a trained model with a constant cardinality of matched FLOPs, retrain from scratch, and see whether accuracy falls.

Watch

Extended reading notes

Core claim

The paper claims that the optimal grouping strategy is layer-dependent and learnable, and that replacing a fixed cardinality like ResNeXt's 32 groups with learned per-layer group counts improves accuracy at matched or lower cost. In their ImageNet experiments, G-ResNeXt101 with budget $b=32$ reaches 79.9% top-1 accuracy versus 79.1% for their re-implemented ResNeXt101 32x4d, with slightly fewer parameters and roughly 0.70x the group-convolution FLOPs. When the learned group numbers are copied back into a standard ResNeXt and trained from scratch, accuracy stays nearly identical (79.8%), showing that the learned grouping strategy, not DGConv's masking, carries the gain. Under tighter budgets the method still holds accuracy: at 0.22x group-convolution FLOPs, G-ResNeXt101 reaches 79.5%. The paper also reports that lower layers tend to learn fewer groups, i.e., more capacity, which it reads as evidence that early layers carry more representational weight.

Load-bearing premise

The load-bearing premise is that the straight-through estimator used to train the binary gates genuinely guides them to useful grouping structures, rather than merely saturating to signs chosen by initialization and gradient noise.

Editorial extensions

If this is right

  • Per-layer cardinality becomes a tunable architectural choice: at a fixed parameter or FLOP budget, a network can spend capacity where it helps instead of spreading it uniformly.
  • A model trained with DGConv can be converted into an ordinary group-convolution network, so the learned structure ships with no runtime overhead and no custom kernels.
  • The observed pattern that early layers prefer fewer groups (more channels per group) provides a concrete design rule for hand-designed efficient networks.
  • The resource-constrained training with budget $b$ gives a direct control knob for the accuracy-complexity trade-off, with reported retention of accuracy at a fraction of group-convolution cost.

Reading between the lines

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

  • If the learned grouping strategies transfer across datasets, per-layer group counts could be treated as a reusable architectural prior rather than something re-learned per task; the paper does not test cross-dataset transfer.
  • DGConv's grouping is a structured form of sparsity, so it could be combined with channel pruning: the learned group boundaries identify which channel bundles are computationally separable, potentially guiding pruning decisions.
  • The straight-through estimator is one route to binarizing the gates; alternative relaxations such as Gumbel-softmax might find different (possibly better) group structures, which the paper does not explore.
  • The observation that early layers favor fewer groups suggests an efficiency guideline for future architectures: keep high-resolution early layers dense or wide, and concentrate grouping in deeper layers.
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

3 major / 5 minor

Summary. The paper proposes a dynamic grouping convolution (DGConv) that learns per-layer group cardinalities end-to-end. DGConv parameterizes a binary channel-connection mask as a Kronecker product of 2x2 blocks controlled by a small set of continuous gates, uses a straight-through estimator for the sign function, and adds a complexity regularizer. The authors build Groupable-ResNeXt by replacing group convolutions in ResNeXt with DGConv and report ImageNet results: G-ResNeXt50/101 outperform their manual ResNet/ResNeXt counterparts at similar or lower complexity, learned group assignments transfer to standard ResNeXt by retraining from scratch (Table 6), and three runs with different seeds produce comparable accuracy (Table 5).

Significance. If the central claim holds, the paper makes a useful contribution: it replaces a manually fixed hyperparameter (group count) with a cheap differentiable mechanism, shows that the learned per-layer cardinalities transfer to standard group convolution, and demonstrates a favorable complexity-accuracy trade-off. Strengths include the explicit transfer experiment in Table 6, the reproducibility runs in Table 5, the small parameter overhead (log of the channel count), and the systematic sweep over complexity budgets in Table 3. The main weakness is that the experimental design does not yet isolate whether the gain comes from the differentiable learning mechanism or simply from allowing non-uniform group counts; this needs a non-uniform control before the central attribution can be accepted.

major comments (3)
  1. [Section 4, Tables 2, 5, and 6] The central claim that DGConv's differentiable learning discovers better group numbers is not fully supported, because every comparison is against ResNeXt with a constant cardinality (32x4d). The authors never test a non-uniform group schedule with the same total FLOPs that is not produced by the differentiable optimization, such as a random assignment of group counts or a hand-designed depth-increasing schedule. Figure 5 indicates that the learned pattern is simple (lower layers use fewer groups), so a hand-craftable or random non-uniform baseline could plausibly match the reported accuracy. Please add at least one such control, trained from scratch with the same complexity budget, and report whether the learned schedules outperform it. Without this control, the improvement could be attributed to relaxing the uniform-cardinality constraint rather than to the learning mechanism.
  2. [Table 3] The G-ResNeXt101 rows for b=96 and b=128 both report 0.22x GConv FLOPs. This is internally inconsistent with the stated budget mapping, where a larger b should correspond to a smaller complexity budget, and it undercuts the complexity-accuracy trade-off conclusions drawn from the table. Please correct the values or explain why two different budgets produce identical FLOPs; also verify the surrounding text statement that b=64 corresponds to 0.5x complexity, which the table lists as 0.47x.
  3. [Section 3.2, Eq. (6)] The optimization of the binary gates relies on a straight-through estimator, but the paper does not analyze how this estimator behaves on the Kronecker-product mask space. The cited reference [5] addresses activation quantization, which is a different optimization setting. Because Figure 6 shows the gates saturating during training, the learned schedules could be influenced by STE dynamics rather than by the loss landscape. Please provide supporting evidence that the differentiable optimization is what drives the gains, for example an ablation with fixed random masks at matched complexity or a comparison with a continuous relaxation of the gates. This directly affects the paper's claim that group counts are "learned" in a meaningful sense.
minor comments (5)
  1. [Section 4, Figure 5 cross-reference] The text says G-ResNeXt50(b=32) and G-ResNeXt101(b=32) correspond to Fig. 5(d) and Fig. 5(a), but the Fig. 5 caption assigns (c) to G-ResNeXt50 b=32 and (d) to G-ResNeXt50 b=96. Please correct the panel reference.
  2. [Section 4 and Tables 2-5] All reported accuracies are single runs except the reproducibility study in Table 5, and no error bars or standard deviations are given for the key comparisons. Please report at least mean values over multiple seeds for the main baselines and the proposed models, or state explicitly that the reproducibility study is the only multi-seed evidence.
  3. [Eq. (9)] Equation (9) writes the total complexity as a sum over ζ_k, while Eq. (8) defines layer-wise terms as ζ_ℓ. Please unify the notation to avoid confusion.
  4. [Throughout] There are several typos and reference-format issues, including "signle" in Section 4, "The Univesity of Hong Kong" in the author affiliation, and the incomplete citation "Yin Penghang et al." for reference [5]. Please proofread and complete the references.
  5. [Section 3.3] The statement that a=0 when ζ≤o and a=α otherwise, with α=-0.02, makes the penalty term [o/ζ]^a discontinuous in the exponent; the paper should clarify how this is implemented in the autodiff computation graph and whether the discontinuity affects training.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: learned group counts are validated on held-out ImageNet and transferred to fresh ResNeXt training.

full rationale

The paper's central claim is that end-to-end learned per-layer group counts improve on a fixed cardinality. The counts are optimized with a classification loss plus a complexity penalty on the training set, and all reported accuracies are measured on the held-out validation set. The strongest anti-circularity control is Table 6, where the learned group counts are extracted, inserted into a standard ResNeXt, and retrained from scratch; the retrained model reaches nearly the same accuracy as the DGConv model. This transfer experiment breaks any reduction by construction: the validation result is not produced by the fitted gates themselves, and the discrete group counts are reused in an independent optimization. The only self-citations, e.g. [17, 18, 26] for optimizing binary variables, are not load-bearing for the empirical improvement and do not justify the central claim. The absence of a random non-uniform group-count baseline is a legitimate experimental-design concern about attribution, but it is not circularity: no prediction is definitionally equal to an input, and no fitted parameter is renamed as a prediction. The STE convergence assumption is a correctness risk, not a circular step.

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

The central result rests on three fitted or hand-chosen hyperparameters (alpha = -0.02, budget scale b, gate initialization), several domain assumptions (square masks, mask-count as complexity proxy, STE validity), and one design choice (2x2 Kronecker factors). No new physical entities are introduced.

free parameters (3)
  • alpha (complexity penalty exponent) = -0.02
    Exponent a in (o/zeta)^a when zeta > o; the paper says MnasNet values (-1, -0.07) were too sensitive, so -0.02 was chosen by hand. It directly sets the trade-off between accuracy and FLOPs.
  • budget scale b = 32, 48, 64, 96, 128, 256, 2 across experiments
    Defines the complexity budget o = sum C_l^2 / b for the whole network; each experiment fixes b and therefore the target FLOPs, making b a fitted or selected hyperparameter.
  • gate initialization = plus or minus 1e-8 random
    Continuous gates are initialized near zero with random sign; different seeds produce slightly different group structures (R2/R3 runs), so the initialization is a free parameter affecting the outcome.
assumptions (5)
  • standard math Kronecker product algebra and the row-sum identity sum of each row of U = product_k (1 + g_k) (Eq. 9)
    Used to derive the complexity formula zeta; standard linear algebra, assumed valid.
  • domain assumption The number of non-zero entries of U measures computational cost (Eq. 8)
    The resource constraint o is defined through this proxy; it ignores memory access and implementation overheads, and would need validation for different hardware.
  • domain assumption C_in equals C_out for the DGConv layers considered (square masks)
    All derivations and experiments use square convolution layers (e.g. 3x3 convs with equal input and output channels); the rectangular case is deferred to Appendix A, which is not available in this text.
  • ad hoc to paper The straight-through estimator provides correct gradients for sign(g) and converges
    Section 3.2 states that gate params are optimized by Straight-Through Estimator and are guaranteed to converge by [5], but no proof or analysis is given for the Kronecker mask space.
  • ad hoc to paper The 2x2 factorization U_k = g_k 1 + (1 - g_k) I spans a useful family of grouping strategies
    This design choice restricts group counts to powers of two and to a specific nested mask structure; the paper's claims of dynamic and non-adjacent grouping depend on this family being expressive enough.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Differentiable Learning-to-Group Channels via Groupable Convolutional Neural Networks." pith.science (2026). https://pith.science/paper/IDBJH6SN

@misc{pith2026190805867,
  author       = {Pith},
  title        = {Pith review of: Differentiable Learning-to-Group Channels via Groupable Convolutional Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IDBJH6SN}},
  note         = {Machine review of arXiv:1908.05867}
}
read the original abstract

Group convolution, which divides the channels of ConvNets into groups, has achieved impressive improvement over the regular convolution operation. However, existing models, eg. ResNeXt, still suffers from the sub-optimal performance due to manually defining the number of groups as a constant over all of the layers. Toward addressing this issue, we present Groupable ConvNet (GroupNet) built by using a novel dynamic grouping convolution (DGConv) operation, which is able to learn the number of groups in an end-to-end manner. The proposed approach has several appealing benefits. (1) DGConv provides a unified convolution representation and covers many existing convolution operations such as regular dense convolution, group convolution, and depthwise convolution. (2) DGConv is a differentiable and flexible operation which learns to perform various convolutions from training data. (3) GroupNet trained with DGConv learns different number of groups for different convolution layers. Extensive experiments demonstrate that GroupNet outperforms its counterparts such as ResNet and ResNeXt in terms of accuracy and computational complexity. We also present introspection and reproducibility study, for the first time, showing the learning dynamics of training group numbers.

Figures

Figures reproduced from arXiv: 1908.05867 by the authors.

Figure 1
Figure 1. Illustration of different convolution strategies, where the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison of group numbers in ResNeXt and Group [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Illustration of structures with relationship matrix U. The hollow circle and solid black circle indicate ‘0’ and ‘1’ respec￾tively. A matrix of ones(a), identity matrix(b) and block diagonal matrix(d) imply regular convolution, depthwise convolution and group convolution (GConv) respectively. (c) and (e) show Dynamic Grouping Convolution (DGConv) under two non-adjacent group strategies respectively, one with a group… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Comparison of the residual building blocks of [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Learned number of groups for each DGConv layer in Groupable-ResNeXt, including: (a) G-ResNeXt101, [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Learning dynamics of group number and learnable gate [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 15 canonical work pages

  1. [5]

    Understanding straight-through estima- tor in training activation quantized neural nets

    Yin Penghang et al. Understanding straight-through estima- tor in training activation quantized neural nets. 2019

    work page 2019

  2. [1]

    Designing neural network architectures using rein- forcement learning

    Bowen Baker, Otkrist Gupta, Nikhil Naik, and Ramesh Raskar. Designing neural network architectures using rein- forcement learning. arXiv preprint arXiv:1611.02167, 2016

    arXiv 2016

  3. [2]

    A constructive arbitrary- degree kronecker product decomposition of tensors

    Kim Batselier and Ngai Wong. A constructive arbitrary- degree kronecker product decomposition of tensors. Numer- ical Linear Algebra with Applications, 24(5):e2097, 2017

    work page 2017

  4. [3]

    Reinforcement learning for architecture search by network transformation

    Han Cai, Tianyao Chen, Weinan Zhang, Yong Yu, and Jun Wang. Reinforcement learning for architecture search by network transformation. arXiv preprint arXiv:1707.04873 , 2017

    arXiv 2017

  5. [4]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009

    2009

  6. [6]

    Delving deep into rectifiers: Surpassing human-level perfor- mance on imagenet classification

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Delving deep into rectifiers: Surpassing human-level perfor- mance on imagenet classification. In Proceedings of the IEEE international conference on computer vision , pages 1026–1034, 2015

    work page 2015

  7. [7]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

    2016

  8. [8]

    Identity mappings in deep residual networks

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Identity mappings in deep residual networks. In European conference on computer vision , pages 630–645. Springer, 2016

    work page 2016

Show all 38 references
  1. [9]

    Mobilenets: Efficient convolu- tional neural networks for mobile vision applications

    Andrew G Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco An- dreetto, and Hartwig Adam. Mobilenets: Efficient convolu- tional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861, 2017

    2017 arXiv

  2. [10]

    Condensenet: An efficient densenet us- ing learned group convolutions

    Gao Huang, Shichen Liu, Laurens Van der Maaten, and Kil- ian Q Weinberger. Condensenet: An efficient densenet us- ing learned group convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 2752–2761, 2018

    2018

  3. [11]

    Batch normalization: Accelerating deep network training by reducing internal co- variate shift

    Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep network training by reducing internal co- variate shift. arXiv preprint arXiv:1502.03167, 2015

    2015 arXiv

  4. [12]

    Imagenet classification with deep convolutional neural net- works

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural net- works. In Advances in neural information processing sys- tems, pages 1097–1105, 2012

    2012

  5. [13]

    Progressive neural architecture search

    Chenxi Liu, Barret Zoph, Maxim Neumann, Jonathon Shlens, Wei Hua, Li-Jia Li, Li Fei-Fei, Alan Yuille, Jonathan Huang, and Kevin Murphy. Progressive neural architecture search. In Proceedings of the European Conference on Com- puter Vision (ECCV), pages 19–34, 2018

    2018

  6. [14]

    Hierarchical repre- sentations for efficient architecture search

    Hanxiao Liu, Karen Simonyan, Oriol Vinyals, Chrisantha Fernando, and Koray Kavukcuoglu. Hierarchical repre- sentations for efficient architecture search. arXiv preprint arXiv:1711.00436, 2017

    2017 arXiv

  7. [15]

    Darts: Differentiable architecture search

    Hanxiao Liu, Karen Simonyan, and Yiming Yang. Darts: Differentiable architecture search. arXiv preprint arXiv:1806.09055, 2018

    2018 arXiv

  8. [16]

    Sgdr: Stochas- tic gradient descent with warm restarts

    Ilya Loshchilov and Frank Hutter. Sgdr: Stochas- tic gradient descent with warm restarts. arXiv preprint arXiv:1608.03983, 2016

    2016 arXiv

  9. [17]

    Switchable normalization for learning-to- normalize deep representation

    Ping Luo, Ruimao Zhang, Jiamin Ren, Zhanglin Peng, and Jingyu Li. Switchable normalization for learning-to- normalize deep representation. IEEE Transactions on Pat- tern Analysis and Machine Intelligence, 2019

    2019

  10. [18]

    Differentiable dynamic nor- malization for learning deep representation

    Ping Luo, Peng Zhanglin, Shao Wenqi, Zhang Ruimao, Ren Jiamin, and Wu Lingyun. Differentiable dynamic nor- malization for learning deep representation. In Kamalika Chaudhuri and Ruslan Salakhutdinov, editors, Proceedings of the 36th International Conference on Machine Learning...

  11. [19]

    Shufflenet v2: Practical guidelines for efficient cnn architec- ture design

    Ningning Ma, Xiangyu Zhang, Hai-Tao Zheng, and Jian Sun. Shufflenet v2: Practical guidelines for efficient cnn architec- ture design. In Proceedings of the European Conference on Computer Vision (ECCV), pages 116–131, 2018

    2018

  12. [20]

    Evolving deep neural networks

    Risto Miikkulainen, Jason Liang, Elliot Meyerson, Aditya Rawal, Daniel Fink, Olivier Francon, Bala Raju, Hormoz Shahrzad, Arshak Navruzyan, Nigel Duffy, et al. Evolving deep neural networks. In Artificial Intelligence in the Age of Neural Networks and Brain Computing , pages 29...

    2019

  13. [21]

    Efficient neural architecture search via parameter sharing

    Hieu Pham, Melody Y Guan, Barret Zoph, Quoc V Le, and Jeff Dean. Efficient neural architecture search via parameter sharing. arXiv preprint arXiv:1802.03268, 2018

    2018 arXiv

  14. [22]

    Xnor-net: Imagenet classification using bi- nary convolutional neural networks

    Mohammad Rastegari, Vicente Ordonez, Joseph Redmon, and Ali Farhadi. Xnor-net: Imagenet classification using bi- nary convolutional neural networks. InEuropean Conference on Computer Vision, pages 525–542. Springer, 2016

    2016

  15. [23]

    Regularized evolution for image classifier architecture search

    Esteban Real, Alok Aggarwal, Yanping Huang, and Quoc V Le. Regularized evolution for image classifier architecture search. arXiv preprint arXiv:1802.01548, 2018

    2018 arXiv

  16. [24]

    Large-scale evolution of image classifiers

    Esteban Real, Sherry Moore, Andrew Selle, Saurabh Sax- ena, Yutaka Leon Suematsu, Jie Tan, Quoc V Le, and Alexey Kurakin. Large-scale evolution of image classifiers. In Pro- ceedings of the 34th International Conference on Machine Learning-Volume 70, pages 2902–2911. JMLR. org, 2017

    2017

  17. [25]

    Mobilenetv2: Inverted residuals and linear bottlenecks

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zh- moginov, and Liang-Chieh Chen. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 4510–4520, 2018

    2018

  18. [26]

    Ssn: Learning sparse switchable normalization via sparsestmax

    Wenqi Shao, Tianjian Meng, Jingyu Li, Ruimao Zhang, Yu- dian Li, Xiaogang Wang, and Ping Luo. Ssn: Learning sparse switchable normalization via sparsestmax. In Pro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 443–451, 2019

    2019

  19. [27]

    Inception-v4, inception-resnet and the impact of residual connections on learning

    Christian Szegedy, Sergey Ioffe, Vincent Vanhoucke, and Alexander A Alemi. Inception-v4, inception-resnet and the impact of residual connections on learning. In Thirty-First AAAI Conference on Artificial Intelligence, 2017

    2017

  20. [28]

    Going deeper with convolutions

    Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1–9, 2015

    2015

  21. [29]

    Rethinking the inception archi- tecture for computer vision

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. Rethinking the inception archi- tecture for computer vision. In Proceedings of the IEEE con- ference on computer vision and pattern recognition , pages 2818–2826, 2016

    2016

  22. [30]

    Mnasnet: Platform-aware neural architec- ture search for mobile

    Mingxing Tan, Bo Chen, Ruoming Pang, Vijay Vasudevan, and Quoc V Le. Mnasnet: Platform-aware neural architec- ture search for mobile. arXiv preprint arXiv:1807.11626 , 2018

    2018 arXiv

  23. [31]

    Fully learnable group convolution for acceleration of deep neural networks

    Xijun Wang, Meina Kan, Shiguang Shan, and Xilin Chen. Fully learnable group convolution for acceleration of deep neural networks. In The IEEE Conference on Computer Vi- sion and Pattern Recognition (CVPR), June 2019

    2019

  24. [32]

    Genetic cnn

    Lingxi Xie and Alan Yuille. Genetic cnn. In Proceedings of the IEEE International Conference on Computer Vision , pages 1379–1388, 2017

    2017

  25. [33]

    Aggregated residual transformations for deep neural networks

    Saining Xie, Ross Girshick, Piotr Doll ´ar, Zhuowen Tu, and Kaiming He. Aggregated residual transformations for deep neural networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 1492– 1500, 2017

    2017

  26. [34]

    Snas: stochastic neural architecture search

    Sirui Xie, Hehui Zheng, Chunxiao Liu, and Liang Lin. Snas: stochastic neural architecture search. arXiv preprint arXiv:1812.09926, 2018

    2018 arXiv

  27. [35]

    In- terleaved group convolutions

    Ting Zhang, Guo-Jun Qi, Bin Xiao, and Jingdong Wang. In- terleaved group convolutions. In Proceedings of the IEEE International Conference on Computer Vision, pages 4373– 4382, 2017

    2017

  28. [36]

    Shufflenet: An extremely efficient convolutional neural net- work for mobile devices

    Xiangyu Zhang, Xinyu Zhou, Mengxiao Lin, and Jian Sun. Shufflenet: An extremely efficient convolutional neural net- work for mobile devices. In Proceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition, pages 6848–6856, 2018

    2018

  29. [37]

    Neural architecture search with reinforcement learning

    Barret Zoph and Quoc V Le. Neural architecture search with reinforcement learning. arXiv preprint arXiv:1611.01578 , 2016

    2016 arXiv

  30. [38]

    Learning transferable architectures for scalable image recognition

    Barret Zoph, Vijay Vasudevan, Jonathon Shlens, and Quoc V Le. Learning transferable architectures for scalable image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 8697–8710, 2018

    2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.