Pith. sign in

REVIEW 5 major objections 6 minor 36 references

Ghost-Connect Net: A Generalization-Enhanced Guidance For Sparse Deep Networks Under Distribution Shifts

T0 review · 5 major / 6 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 This paper claims that pruning guided by a companion network built from cross-layer activation correlations improves sparse network accuracy under distribution shift.

desk verdict The companion-network idea is genuinely new and the experimental sweep is thorough, but the theory section doesn't justify the implemented method and the reported gains are too small and inconsistent to support the central claims. read the letter →

arxiv 2411.09199 v1 pith:RYVAFZ3I submitted 2024-11-14 cs.LG

Mary Isabelle Wisell , Salimeh Yasaei Sekeh This is my paper · ORCID
classification cs.LG
keywords deepneuralnetworkpruningdistributionshiftconnectivitymatrixPearsoncorrelationcompanionsparsenetworksgeneralizationCIFAR-10
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 claims that pruning a deep network is safer under distribution shift when the pruning decisions are guided by a companion network that records how strongly the activation states of consecutive layers are correlated. This companion, GC-Net, is built once from the training data, pruned with any of several standard methods, and its pruned locations are mapped back onto the original network. The authors report that hybrid pruning, letting GC-Net guide later layers while pruning earlier layers directly, matches or beats direct pruning on CIFAR-10, Fashion MNIST, and Tiny ImageNet under color-jitter, noise-and-blur, and lighting-occlusion shifts. They also give a theoretical argument that the optimal mask maximizes a score built from adjacent weight magnitudes, which they read as a tight link between connectivity-guided and magnitude-based pruning. If the claim holds, sparse networks could keep their efficiency advantage without sacrificing robustness when deployment data differs from training data.

What carries the argument

The load-bearing object is the connectivity matrix $R$ defined as the Pearson correlation between activation states of consecutive layers, $\rho(f_i^{(l)}, f_j^{(l+1)})$, computed on the training distribution and reshaped into the weights of GC-Net. GC-Net is a copy of the original architecture with an identity first layer; its weights are the expanded $R$ matrices, pruned with L1, L2, C-SNIP, or OS-SynFlow, and the resulting masks are mapped back onto the original network. For skip connections the two incoming connectivity matrices are added, and for VGG pooling layers the matrix is transposed and duplicated along the pooling kernel. The theoretical chain uses a concatenation operation $\Delta$ on information flow to bound the change in downstream connectivity, ultimately producing the maximization problem in Eq. (29) with scores $g^{k-1,k}=(|\omega^{k-1}\omega^k|)^\top s^L$, which the authors identify as the tight link to magnitude pruning.

What would settle it

Construct a shifted dataset that keeps each feature's marginal distribution but reorders cross-layer activation correlations, then run the GC-Net pipeline and compare against direct magnitude pruning; if shifted accuracy is not higher, the transfer premise fails. Separately, evaluate the inequality in Eq. (18) on a trained network: for a weight with $|\omega|<1/C_y$, the claimed bound $(1+C_y|\omega|)\le C|\omega|$ is false, so the proof chain would need a lower bound on weight magnitude.

Watch

Extended reading notes

Core claim

On its own terms, the central claim is that a network pruned with GC-Net guidance, particularly on its later layers, maintains or improves accuracy on shifted distributions compared with pruning the same architecture directly. GC-Net is a copy of the original network whose weights are the Pearson correlations between activation states of consecutive layers, expanded to match layer dimensions; pruning GC-Net and copying the masks back therefore removes connections that are weakly correlated across layers rather than merely small in magnitude. The optimization section derives that the pruning objective reduces to maximizing $\sum_i g_i \tilde{s}_i$, where $g^{k-1,k} = (|\omega^{k-1}\omega^k|)^\top s^L$, so the guidance implicitly uses products of adjacent weight magnitudes, information that standard magnitude pruning ignores. Experimentally, the paper reports that the back-half and last-25% hybrid variants are the strongest, with gains most visible at 20% sparsity and on the color-jitter and noise-and-blur shifts.

Load-bearing premise

The central assumption is that activation correlations measured once on the training distribution still mark the connections that matter after a distribution shift; the theoretical bound additionally assumes an inequality that does not hold for weights near zero.

Editorial extensions

If this is right

  • A sparse network pruned with GC-Net guidance on its later layers can match or exceed the shifted-distribution accuracy of the same network pruned directly, with the clearest gains at 20% sparsity on the color-jitter and noise-and-blur shifts.
  • Because the optimal mask in Eq. (29) depends on $|\omega^{k-1}\omega^k|$ rather than $|\omega|$ alone, GC-Net guidance incorporates magnitude information that spans adjacent layers, which plain magnitude pruning does not use.
  • GC-Net is created once from the training distribution and is not used at inference, so the extra FLOPs are a one-time cost concentrated in connectivity-matrix computation.
  • The best hybrid placement is architecture- and dataset-dependent: the back-half and last-25% variants are the strongest for ResNet-18 and VGG16-BN, and the connectivity metric (Pearson correlation or cosine similarity) can be swapped with only small accuracy changes.
  • The guidance mechanism transfers across pruning methods, since gains appear with L1, L2, C-SNIP, and OS-SynFlow rather than being tied to a single pruning rule.

Reading between the lines

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

  • The paper leaves implicit that the Eq. (29) link suggests a parameter-free pruning rule of the form $(|\omega^{k-1}\omega^k|)^\top s^L$, which could be evaluated without building a full companion network; testing that rule directly would isolate whether the connectivity matrices add anything beyond adjacent weight products.
  • The transfer premise could be probed directly by building GC-Net on the shifted distribution and comparing masks; if the optimal masks change substantially, the method's gains depend on correlation stability, which the paper does not measure.
  • The reported CIFAR-10 gains are frequently within a few tenths of a percent, so the practical value may rest on high sparsity levels or particular shift types rather than being universal; a reader should check the appendix tables for the settings where the hybrid wins.
  • A natural extension is to recompute GC-Net periodically during fine-tuning or at test time, something the authors list as future work; the static one-time computation is a deliberate design choice, not a necessity.
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

5 major / 6 minor

Summary. The paper introduces Ghost-Connect Net (GC-Net), a companion network whose weights encode connectivity scores (Pearson correlations between activation states of consecutive layers) of a pre-trained original network. The pruning process prunes GC-Net and maps the pruned indices back to the original network, optionally combined with direct magnitude-based pruning on a subset of layers (hybrid variants). Experiments on CIFAR-10, Fashion MNIST, and Tiny ImageNet under three synthetic distribution shifts (CJG, RNB, LO) report accuracies for several sparsity levels and pruning methods. Section 5 claims a theoretical foundation: a bound on a surrogate loss leading to an optimal mask that maximizes a weight-product score, which the authors argue establishes a 'tight connection' between GC-Net and magnitude pruning.

Significance. The problem of making pruned networks robust to distribution shifts is timely and relevant, and the idea of using a separate connectivity-monitoring network to guide pruning is a creative contribution. The paper includes a large set of experiments and a FLOPs analysis, which are strengths. However, the theoretical section contains a false inequality and an unconstrained optimization whose solution is trivial, and the derivation uses a different connectivity notion than the implemented Pearson correlation. The empirical results show only marginal and inconsistent differences relative to the baseline, with no statistical validation. As presented, the central claims of theoretical grounding and empirical improvement are not supported. If the theory were repaired and the experiments strengthened with standard benchmarks and significance tests, the underlying idea could be worth pursuing.

major comments (5)
  1. [Section 5, Eq. (18)] The inequality (1 + C_y |ω^{(k)}|) ≤ C |ω^{(k)}| cannot hold for any constant C when |ω^{(k)}| is sufficiently small, because the left-hand side is bounded below by 1 while the right-hand side tends to 0; this step is load-bearing because it converts the bound into the product |ω^{(k-1)}ω^{(k)}| used in Eq. (20) and Eq. (29), so the theoretical derivation collapses.
  2. [Section 5, Eq. (7) vs Algorithm 2] The derivation takes φ to be dot product and ρ to be 'uniform function', but the implemented GC-Net computes connectivity via Pearson correlation (Algorithm 2, Step 2). The paper does not establish any relationship between the derived bound and Pearson-based connectivity, so the theoretical foundation does not apply to the method evaluated in Section 4.
  3. [Section 5, Eq. (29)] The maximization argmax_{s̃} Σ_i g_i s̃_i with s̃_i ∈ [0,1] and g_i ≥ 0 has the trivial solution s̃_i = 1 for all i, i.e., no pruning. Since no sparsity constraint is imposed, the 'optimal mask' prunes nothing, and the claim that this captures filter importance is vacuous.
  4. [Section 4, Tables 1-42] Across the reported settings, the differences between GC-Net hybrids and the baseline pruning method are mostly within 0.3 percentage points and inconsistent in direction (e.g., Table 1, L1-Norm on CJG: GC-Net-B25% 73.32 vs Original 74.38; C-SNIP on RNB: GC-Net-B25% 24.51 vs Original 24.62). With only 3 trials and no significance tests or confidence intervals, these results do not support the claim that GC-Net improves robustness under distribution shifts.
  5. [Section 4.1] The distribution shifts CJG, RNB, and LO are custom synthetic perturbations, and the evaluation does not include comparisons to state-of-the-art robust pruning or domain generalization methods cited in Related Work. This limits the external validity of the empirical conclusions.
minor comments (6)
  1. [Eq. (3)] The simplification ρ = E[f^{(l)}f^{(l+1)}|Y] requires the filters to be zero-mean and unit-variance; this should be presented as an explicit normalization assumption rather than a general equality.
  2. [Eq. (6)] The expression has a mismatched parenthesis; it should read φ(s^{(L)}, |ρ(G^{(l+1,L)}(x)) - ρ(G^{(l+1,L)}(s̃^{(l)} ⊙ x))|).
  3. [Fig. 3] The label 'CG-Net - FH' should be 'GC-Net - FH'.
  4. [Algorithm 1] The hyperparameter K is used to define the hybrid pruning set but is not defined in the list at the beginning of the algorithm; clarify its role.
  5. [Section 5, Eq. (9)] The ∆ operation is described as matrix concatenation, but Eq. (9) applies it to scalar quantities; the intended algebraic structure should be defined precisely.
  6. [Reference [28]] Reference [28] appears to be a proceedings title rather than a standard citation; please provide the full author and paper details.

Circularity Check

0 steps flagged · score 2.0 of 10

No circularity in the empirical method; only a minor background self-citation, while the theoretical gap is a correctness issue rather than a circular reduction.

full rationale

The empirical pipeline is self-contained: GC-Net connectivity matrices are computed once from activations on the training distribution D1 (Algorithm 2, Section 8.3), masks are produced by external pruning methods, and the pruned network is fine-tuned on D1 and tested on D2. No accuracy constant or mask is fitted to D2, and the benchmarks (CIFAR-10, FMNIST, Tiny-IN with synthetic shifts) are external, so the central empirical claim does not reduce to its inputs. The theoretical section (Eqs. 5-29) attempts to bound a connectivity-based loss and arrives at a weight-product importance score; although it is mathematically disconnected from the implemented Pearson-correlation procedure and contains invalid steps (e.g., Eq. 18 assumes 1 + C_y|omega| <= C|omega|, which fails for small weights, and Eq. 29 without a sparsity constraint is maximized by keeping all connections), those are correctness failures, not circularity: the conclusion is not identical to an assumption by construction. The only self-citation is ref. [3] for the phrase "so called information flow" in Eq. (1); the same Pearson-correlation quantity is independently defined in Eq. (3) and used in Section 5, so the citation is background terminology, not a load-bearing justification. Score 2 reflects this one minor non-load-bearing self-citation; the central claim retains independent content.

Assumptions & free parameters 3 free parameters · 5 assumptions · 1 invented entities

The paper does not fit model parameters to target accuracies, but the method depends on hand-chosen hyperparameters (sparsity level, hybrid region, connectivity metric). The theoretical section relies on several unproven or invalid assumptions, most notably the false inequality in Eq. (18). GC-Net itself is an invented algorithmic entity without independent evidence outside the paper's experiments.

free parameters (3)
  • Sparsity level α = 20%, 40%, 60%, 80%
    Chosen by hand per experiment; results vary strongly with α, and the main text highlights only the 20% sparsity cases.
  • Hybrid layer region K = Full, FH, BH, B25%
    Four configurations selecting which layers are pruned via GC-Net; the main text reports only the best two (BH, B25%) as top-performing hybrids.
  • Connectivity metric = Pearson correlation or cosine similarity
    Two variants tested in Section 4.4; the choice affects results and is not theoretically justified.
assumptions (5)
  • domain assumption Activation function σ is bounded
    Assumed in Section 3.1 and used in Section 5 to bound activation values and derive Lipschitz-style inequalities.
  • domain assumption Filters are normalized (zero mean and unit variance)
    Stated in Section 3.1 (Eq. 3) to simplify connectivity to a conditional expectation; needed for the theoretical simplification.
  • ad hoc to paper φ satisfies the subadditivity inequalities in Assumption 1
    Assumption 1 in Section 5 (Eq. 23) is introduced solely to push the bound through; it is not justified for the dot product φ used in Eq. (27).
  • ad hoc to paper There exists a constant C with (1 + C_y |ω^(k)|) ≤ C |ω^(k)|
    Invoked in Eq. (18) to complete the bound; this is false for weights near zero and breaks the derivation.
  • domain assumption Activation functions are Lipschitz continuous and filter values are bounded
    Used in Section 5 (Eqs. 15-17) to obtain the Lipschitz bounds on filter outputs.
invented entities (1)
  • Ghost-Connect Net (GC-Net)
    purpose: A companion network whose weights are inter-layer activation connectivity matrices, pruned to derive the mask mapped back to the original network.
    It is a methodological construct rather than a physical entity; its only evidence is the paper's own experiments, and no external falsifiable prediction is made.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Ghost-Connect Net: A Generalization-Enhanced Guidance For Sparse Deep Networks Under Distribution Shifts." pith.science (2026). https://pith.science/paper/RYVAFZ3I

@misc{pith2026241109199,
  author       = {Pith},
  title        = {Pith review of: Ghost-Connect Net: A Generalization-Enhanced Guidance For Sparse Deep Networks Under Distribution Shifts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RYVAFZ3I}},
  note         = {Machine review of arXiv:2411.09199}
}
read the original abstract

Sparse deep neural networks (DNNs) excel in real-world applications like robotics and computer vision, by reducing computational demands that hinder usability. However, recent studies aim to boost DNN efficiency by trimming redundant neurons or filters based on task relevance, but neglect their adaptability to distribution shifts. We aim to enhance these existing techniques by introducing a companion network, Ghost Connect-Net (GC-Net), to monitor the connections in the original network with distribution generalization advantage. GC-Net's weights represent connectivity measurements between consecutive layers of the original network. After pruning GC-Net, the pruned locations are mapped back to the original network as pruned connections, allowing for the combination of magnitude and connectivity-based pruning methods. Experimental results using common DNN benchmarks, such as CIFAR-10, Fashion MNIST, and Tiny ImageNet show promising results for hybridizing the method, and using GC-Net guidance for later layers of a network and direct pruning on earlier layers. We provide theoretical foundations for GC-Net's approach to improving generalization under distribution shifts.

Figures

Figures reproduced from arXiv: 2411.09199 by the authors.

Figure 1
Figure 1. GC-Net Guidance Overview: (Blue module) Generate activation matrices (step 1), create connectivity matrices via Pearson Correlation (step 2), expand matrices as GC-Net weights (step 3). (Red module) Prune GC-Net. (Green module) Map pruned GC-Net to original network. (Yellow module) Prune remaining original network layers [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. When the outputs of two layers converge into a single [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Comparison of ResNet-18 OS-SynFlow FLOPs for original and hybrid methods using different similarity metrics for CIFAR-10. Pruning Method Hybrid Type CIFAR-10 DS Dataset Orig(Acc1) CJG RNB LO L1-Norm AccO: 93.14% GC-Net-BH 93.05 73.52 23.63 13.82 GC-Net-B25% 93.11 73.32 24.24 13.75 Original 93.04 74.38 24.49 13.69 L2-Norm AccO: 93.14% GC-Net-BH 91.76 70.89 23.23 13.74 GC-Net-B25% 91.85 70.63 23.46 14.15 Original 91.9… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 24 canonical work pages

  1. [1]

    How does loss function affect generalization per- formance of deep learning? application to human age esti- mation

    Ali Akbari, Muhammad Awais, Manijeh Bashar, and Josef Kittler. How does loss function affect generalization per- formance of deep learning? application to human age esti- mation. In International Conference on Machine Learning , pages 141–151. PMLR, 2021

    work page 2021

  2. [2]

    Rage: Robust age estimation through subject anchoring with consistency regularisation

    Ali Akbari, Muhammad Awais, Soroush Fatemifar, Syed Safwan Khalid, and Josef Kittler. Rage: Robust age estimation through subject anchoring with consistency regularisation. IEEE transactions on pattern analysis and machine intelligence, 46(3):1603–1617, 2022

    work page 2022

  3. [3]

    Theoretical un- derstanding of the information flow on continual learning performance

    Joshua Andle and Salimeh Yasaei Sekeh. Theoretical un- derstanding of the information flow on continual learning performance. In European Conference on Computer Vision, pages 86–101. Springer, 2022

    work page 2022

  4. [4]

    Invariant risk minimization

    Martin Arjovsky, L ´eon Bottou, Ishaan Gulrajani, and David Lopez-Paz. Invariant risk minimization. arXiv preprint arXiv:1907.02893, 2019

    arXiv 1907

  5. [5]

    Deeppicar: A low-cost deep neural network- based autonomous car

    Michael G Bechtel, Elise McEllhiney, Minje Kim, and Heechul Yun. Deeppicar: A low-cost deep neural network- based autonomous car. In2018 IEEE 24th international con- ference on embedded and real-time computing systems and applications (RTCSA), pages 11–21. IEEE, 2018

    work page 2018

  6. [6]

    The elastic lottery ticket hypothesis

    Xiaohan Chen, Yu Cheng, Shuohang Wang, Zhe Gan, Jingjing Liu, and Zhangyang Wang. The elastic lottery ticket hypothesis. Advances in Neural Information Processing Sys- tems, 34:26609–26621, 2021

    work page 2021

  7. [7]

    Urban artificial intelligence: From au- tomation to autonomy in the smart city

    Federico Cugurullo. Urban artificial intelligence: From au- tomation to autonomy in the smart city. Frontiers in Sustain- able Cities, 2:38, 2020

    work page 2020

  8. [8]

    A winning hand: Compress- ing deep networks can improve out-of-distribution robust- ness

    James Diffenderfer, Brian Bartoldson, Shreya Chaganti, Jize Zhang, and Bhavya Kailkhura. A winning hand: Compress- ing deep networks can improve out-of-distribution robust- ness. Advances in neural information processing systems , 34:664–676, 2021

    work page 2021

Show all 36 references
  1. [9]

    Depgraph: Towards any structural pruning

    Gongfan Fang, Xinyin Ma, Mingli Song, Michael Bi Mi, and Xinchao Wang. Depgraph: Towards any structural pruning. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 16091–16101, 2023

    2023

  2. [10]

    Pruning neural networks at ini- tialization: Why are we missing the mark? arXiv preprint arXiv:2009.08576, 2020

    Jonathan Frankle, Gintare Karolina Dziugaite, Daniel M Roy, and Michael Carbin. Pruning neural networks at ini- tialization: Why are we missing the mark? arXiv preprint arXiv:2009.08576, 2020

    2009 arXiv

  3. [11]

    Mit ad- vanced vehicle technology study: Large-scale naturalistic driving study of driver behavior and interaction with automa- tion

    Lex Fridman, Daniel E Brown, Michael Glazer, William An- gell, Spencer Dodd, Benedikt Jenik, Jack Terwilliger, Alek- sandr Patsekin, Julia Kindelsberger, Li Ding, et al. Mit ad- vanced vehicle technology study: Large-scale naturalistic driving study of driver behavior and inte...

    2019

  4. [12]

    The state of sparsity in deep neural networks

    Trevor Gale, Erich Elsen, and Sara Hooker. The state of sparsity in deep neural networks. arXiv preprint arXiv:1902.09574, 2019

    1902 arXiv

  5. [13]

    A unified view of label shift estimation.Ad- vances in Neural Information Processing Systems, 33:3290– 3300, 2020

    Saurabh Garg, Yifan Wu, Sivaraman Balakrishnan, and Zachary Lipton. A unified view of label shift estimation.Ad- vances in Neural Information Processing Systems, 33:3290– 3300, 2020

    2020

  6. [14]

    A survey of uncertainty in deep neural networks

    Jakob Gawlikowski, Cedrique Rovile Njieutcheu Tassi, Mohsin Ali, Jongseok Lee, Matthias Humt, Jianxiang Feng, Anna Kruspe, Rudolph Triebel, Peter Jung, Ribana Roscher, et al. A survey of uncertainty in deep neural networks. Arti- ficial Intelligence Review, 56(Suppl 1):1513–1589, 2023

    2023

  7. [15]

    Ghostnet: More features from cheap operations

    Kai Han, Yunhe Wang, Qi Tian, Jianyuan Guo, Chunjing Xu, and Chang Xu. Ghostnet: More features from cheap operations. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1580–1589, 2020

    2020

  8. [16]

    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

  9. [17]

    Benchmarking neu- ral network robustness to common corruptions and perturba- tions

    Dan Hendrycks and Thomas Dietterich. Benchmarking neu- ral network robustness to common corruptions and perturba- tions. In International Conference on Learning Representa- tions, 2018

    2018

  10. [18]

    What do compressed deep 20 neural networks forget? arXiv preprint arXiv:1911.05248, 2019

    Sara Hooker, Aaron Courville, Gregory Clark, Yann Dauphin, and Andrea Frome. What do compressed deep 20 neural networks forget? arXiv preprint arXiv:1911.05248, 2019

    1911 arXiv

  11. [19]

    Wilds: A benchmark of in-the- wild distribution shifts

    Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubra- mani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, et al. Wilds: A benchmark of in-the- wild distribution shifts. In International conference on machine learning...

    2021

  12. [20]

    Out-of-distribution general- ization via risk extrapolation (rex)

    David Krueger, Ethan Caballero, Joern-Henrik Jacobsen, Amy Zhang, Jonathan Binas, Dinghuai Zhang, Remi Le Priol, and Aaron Courville. Out-of-distribution general- ization via risk extrapolation (rex). In International confer- ence on machine learning, pages 5815–5826. PMLR, 2021

    2021

  13. [21]

    Fine-tuning can distort pretrained fea- tures and underperform out-of-distribution

    Ananya Kumar, Aditi Raghunathan, Robbie Jones, Tengyu Ma, and Percy Liang. Fine-tuning can distort pretrained fea- tures and underperform out-of-distribution. arXiv preprint arXiv:2202.10054, 2022

    2022 arXiv

  14. [22]

    Layer-adaptive sparsity for the magnitude-based pruning

    Jaeho Lee, Sejun Park, Sangwoo Mo, Sungsoo Ahn, and Jin- woo Shin. Layer-adaptive sparsity for the magnitude-based pruning. arXiv preprint arXiv:2010.07611, 2020

    2010 arXiv

  15. [23]

    Snip: Single-shot network pruning based on connec- tion sensitivity

    Namhoon Lee, Thalaiyasingam Ajanthan, and Philip HS Torr. Snip: Single-shot network pruning based on connec- tion sensitivity. arXiv preprint arXiv:1810.02340, 2018

    2018 arXiv

  16. [24]

    Ea- gleeye: Fast sub-net evaluation for efficient neural network pruning

    Bailin Li, Bowen Wu, Jiang Su, and Guangrun Wang. Ea- gleeye: Fast sub-net evaluation for efficient neural network pruning. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceed- ings, Part II 16, pages 639–654. Springer, 2020

    2020

  17. [25]

    Do we actually need dense over- parameterization? in-time over-parameterization in sparse training

    Shiwei Liu, Lu Yin, Decebal Constantin Mocanu, and Mykola Pechenizkiy. Do we actually need dense over- parameterization? in-time over-parameterization in sparse training. In International Conference on Machine Learning, pages 6989–7000. PMLR, 2021

    2021

  18. [26]

    Towards robust neural networks via random self- ensemble

    Xuanqing Liu, Minhao Cheng, Huan Zhang, and Cho-Jui Hsieh. Towards robust neural networks via random self- ensemble. In Proceedings of the european conference on computer vision (ECCV), pages 369–385, 2018

    2018

  19. [27]

    huyvnphan/pytorch cifar10, January 2021

    Huy Phan. huyvnphan/pytorch cifar10, January 2021

    2021

  20. [28]

    Understanding and Mitigating the Tradeoff between Robustness and Accuracy, 2022

    Proceedings of the 37th International Conference on Ma- chine Learning, PMLR. Understanding and Mitigating the Tradeoff between Robustness and Accuracy, 2022

    2022

  21. [29]

    Very deep convo- lutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convo- lutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014

    2014 arXiv

  22. [30]

    Neural machine translation: A review

    Felix Stahlberg. Neural machine translation: A review. Jour- nal of Artificial Intelligence Research, 69:343–418, 2020

    2020

  23. [31]

    Test-time training with self- supervision for generalization under distribution shifts

    Yu Sun, Xiaolong Wang, Zhuang Liu, John Miller, Alexei Efros, and Moritz Hardt. Test-time training with self- supervision for generalization under distribution shifts. In International conference on machine learning, pages 9229–

  24. [32]

    Pruning neural networks without any data by iter- atively conserving synaptic flow

    Hidenori Tanaka, Daniel Kunin, Daniel L Yamins, and Surya Ganguli. Pruning neural networks without any data by iter- atively conserving synaptic flow. Advances in neural infor- mation processing systems, 33:6377–6389, 2020

    2020

  25. [33]

    Neu- ral pruning via growing regularization

    Huan Wang, Can Qin, Yulun Zhang, and Yun Fu. Neu- ral pruning via growing regularization. arXiv preprint arXiv:2012.09243, 2020

    2012 arXiv

  26. [34]

    Learning structured sparsity in deep neural net- works

    Wei Wen, Chunpeng Wu, Yandan Wang, Yiran Chen, and Hai Li. Learning structured sparsity in deep neural net- works. Advances in neural information processing systems , 29, 2016

    2016

  27. [35]

    User scheduling for het- erogeneous multiuser mimo systems: A subspace viewpoint

    Xinping Yi and Edward KS Au. User scheduling for het- erogeneous multiuser mimo systems: A subspace viewpoint. IEEE Transactions on Vehicular Technology , 60(8):4004– 4013, 2011

    2011

  28. [36]

    A fre- quency pattern mining model based on deep neural network for real-time classification of heart conditions

    Hyun Yoo, Soyoung Han, and Kyungyong Chung. A fre- quency pattern mining model based on deep neural network for real-time classification of heart conditions. In Health- care, volume 8, page 234. MDPI, 2020. 21

    2020

Pith tools

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