Pith. sign in

REVIEW 4 major objections 4 minor 37 references

Knowledge Grafting: A Mechanism for Optimizing AI Model Deployment in Resource-Constrained Environments

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims a weed-detection model can shrink from 64.39 MB to 7.38 MB by grafting selected intermediate features into a small rootstock network, then beat the donor on validation and unseen test data.

desk verdict The grafting metaphor is strained, the math is generic, and the central claim is undercut by a missing baseline and an internal contradiction about which model supplies the scion. read the letter →

arxiv 2507.19261 v1 pith:W6XZJC3Z submitted 2025-07-25 cs.AI cs.LGcs.PF

Osama Almurshed , Ashish Kaushal , Asmail Muftah , Nitin Auluck , Omer Rana This is my paper · ORCID
classification cs.AIcs.LGcs.PF
keywords knowledgegraftingmodelcompressionfeatureselectionedgeAIdeploymentglobalaveragepoolingweeddetectionVGG16DeepWeeds
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

Knowledge grafting is proposed as a way to shrink a large trained network for deployment on devices with limited memory and compute: instead of compressing all of the donor model, the method pulls only a few of its intermediate feature maps (the scion) through global average pooling and connects them by new dense layers to a small rootstock network. The paper's central claim is that this selective transfer can cut the model from 64.39 MB to 7.38 MB (an 88.54% reduction) while improving generalization, with the grafted model reaching 89.97% validation accuracy against the donor's 87.47% and 90.45% accuracy on unseen test data. This matters because it would mean the usual size-versus-accuracy trade-off is not fixed: edge devices, such as agricultural robots doing weed detection, could run accurate models without cloud connections or dedicated AI hardware. The evidence is a single DeepWeeds weed-classification study, plus a mathematical formulation of layer selection as a combinatorial optimization with two dual objectives: size-constrained performance maximization and performance-constrained size minimization.

What carries the argument

The load-bearing mechanism is the grafting union: selected layer outputs $f_i(x)$ from the donor model are reduced with Global Average Pooling, concatenated into one vector $h(x)=C(\{g_i(x)\mid i\in S\})$, and mapped through new dense layers with LeakyReLU/ReLU and dropout to a softmax output, giving the grafted model $G(x)=M(C(\{T(f_i(x))\mid i\in S\}))$. The mathematical framework makes the choice of layers a binary selection vector $s\in\{0,1\}^m$ over the donor's $m$ layers, and poses two dual optimization problems: maximize performance subject to $\mathrm{Size}(G_s)\le \mathrm{Size}_{\max}$, or minimize size subject to $P(G_s)\ge P_{\min}$. In the reported experiment the scion is taken from layers 8, 9, and 10, and the rootstock is trained for 18 epochs with Adam and categorical cross-entropy; the framework itself is what turns grafting from a heuristic into a searchable optimization problem.

What would settle it

Measure the peak memory and inference time of the proposed grafted model on a target edge device when run end-to-end, and compare that with the stated 7.38 MB; if running the donor's layers 8-10 is required to produce the scion features on each input, the peak footprint will exceed the reported size. A second check is an ablation: train the same rootstock architecture without the grafted features and see whether its accuracy collapses, which would confirm that the grafting itself is the source of the improvement.

Watch

Extended reading notes

Core claim

The discovery the paper asserts is that the grafting operation itself does the work: the rootstock inherits the donor's useful mid-level features, from the paper's chosen layers 8, 9, and 10, through global average pooling and concatenation, and therefore starts from a much stronger baseline than a small model trained from scratch. The resulting grafted model, with 1,934,665 parameters versus the donor's 16,880,201, shows a training/validation accuracy gap of only about 1.87 percentage points, whereas the donor overfits with a gap near 10 points; the rootstock's validation loss is 0.2976 against the donor's 0.5068, and on the held-out test set it scores 90.45% accuracy, 92.64% precision, 89.13% recall, and an AUC of 0.9926. The paper reads these numbers as evidence that smaller models built by selective feature transfer can outperform their larger donors on generalization while using about 8.7 times less storage.

Load-bearing premise

The load-bearing premise is that the scion features are usable by the rootstock without including the donor layers that generated them, and if those layers must be loaded and run at inference time to produce the features, the claimed 7.38 MB footprint is not the full model cost.

Editorial extensions

If this is right

  • If the result holds, a pretrained VGG16-class network can be deployed at roughly 11% of its original footprint without sacrificing accuracy, which is enough to make on-device weed detection feasible on agricultural robots with no cloud link.
  • The reported 8.7x size reduction exceeds the paper's literature baselines of 2-4x for quantization, 3-5x for pruning, and 5-6x for knowledge distillation, positioning grafting as a distinct alternative rather than a variant of those methods.
  • The narrow training/validation gap suggests the grafted features act as a regularizer, so the same optimization can be framed either as minimizing size for a required accuracy or as maximizing accuracy for a fixed size budget.
  • Because only selected feature maps are transferred, the method can be composed with quantization and pruning, and with automated search over the selection vector, which are the paper's stated future extensions.

Reading between the lines

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

  • Editorial inference: the paper does not specify whether scion features are precomputed or produced by running the donor's layers at inference; only in the latter case is the 7.38 MB footprint the full deployment cost, so a fair comparison should measure end-to-end peak memory.
  • Editorial inference: the performance gain could come partly from the donor's fine-tuned weights and partly from the new dense classifier; an ablation that trains the same rootstock from random initialization without the scion would separate the two contributions.
  • Editorial inference: because the selection vector is just an index set over layers, grafting can be applied to any paired donor/rootstock pair, including modern transformer backbones, with the main practical constraint being the compatibility of feature-map shapes and pooled dimensions.
  • Editorial inference: the dual-objective formulation suggests a testable extension: for a fixed size budget, search over layer subsets and plot the Pareto frontier of size versus accuracy; the manual choice of layers 8-10 should lie on that frontier if the mechanism is doing the claimed work.
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 / 4 minor

Summary. The manuscript proposes 'knowledge grafting,' a technique that selects intermediate feature maps (the scion) from a large donor model and appends them, after global average pooling and concatenation, to a smaller rootstock model, with the goal of deploying high-accuracy models in resource-constrained environments. The authors formalize the selection as a binary optimization problem with two objectives, describe an agricultural weed-detection use case on the DeepWeeds dataset, and report a drop from 64.39 MB to 7.38 MB with validation accuracy rising from 87.47% to 89.97% and test accuracy of 90.45%. The paper compares these results with state-of-the-art models and with other compression techniques, arguing that grafting outperforms quantization, pruning, and knowledge distillation.

Significance. The idea of transferring intermediate features to a lightweight model is not new in itself, but the horticulturally framed formulation and the claimed simultaneous size reduction and accuracy improvement would be of practical interest for edge deployment if substantiated. The paper is clearly written and the mathematical notation in Section 4 is tidy, but the contribution as presented is not reproducible: no code, data, or trained models are released, no repeated experiments or error bars are given, and the central result relies on a single run. The most important weakness is internal: the description of where the scion comes from is contradictory, and no rootstock-only baseline is reported. These issues are not merely cosmetic; they bear directly on whether the claimed 88.54% size reduction and accuracy gain are real. I nonetheless credit the authors for a useful use-case discussion and for explicitly acknowledging in Section 9 that the current layer selection is manual, which is consistent with my reading that the Section 4 optimization is not exercised.

major comments (4)
  1. [§5 Step 3 vs §2/Fig. 2] Section 5, Step 3 states that 'Specific layers (8, 9, and 10) are selected from the rootstock model,' while Section 2 and Fig. 2 describe the scion as being removed from the larger donor model. These are incompatible. In a VGG16-style network, layers 8–10 correspond to three 512-channel convolutional layers containing on the order of 7 million parameters, which cannot be part of a rootstock that has only 1,934,665 total parameters. If inference uses the donor's feature extractor to compute the scion, then the 7.38 MB footprint omits the parameters actually used; if the layers are part of the rootstock, then no donor features are grafted and the mechanism in Section 2 is not what is tested. Either way, the headline size reduction and the accuracy gain cannot both be attributed to grafting as described.
  2. [§6] The paper compares the grafted rootstock only against the donor model. There is no experiment that trains the rootstock architecture without the scion (e.g., without the features from layers 8–10) under the same protocol. Therefore the observed improvement from 87.47% to 89.97% validation accuracy, and the test accuracy of 90.45%, cannot be attributed to the grafting mechanism rather than to differences in architecture, regularization, or training dynamics. In addition, the donor's test accuracy is never reported: the abstract and Section 6 compare rootstock test accuracy (90.45%) with donor validation accuracy (87.47%), which is not an apples-to-apples comparison.
  3. [§4 and §9] Equations (7)–(8) define two selection objectives, but the experimental selection S = {8, 9, 10} is made by hand, as Section 9 explicitly acknowledges ('Our current implementation relies on manually selecting layers and features with predetermined grafting points'). The combinatorial search space of 2^m is never explored, no heuristic is run, and no value of the objective in (7) or (8) is computed for the chosen configuration. Thus the mathematical framework is decorative with respect to the reported results; the paper does not demonstrate the 'systematic feature selection' promised in the introduction.
  4. [§6.2–§7 and Tables 2–4] The reported numbers come from a single training run with no code, data, or model release, and no error bars, so the 90.45% test accuracy and 0.9926 AUC cannot be assessed for statistical significance. The comparison tables contain internal inconsistencies that undermine confidence: Table 4 lists 'Parameter Reduction' as '90.45%' although Table 2 gives 88.54%; the same table says pruning achieves 'up to 67%' parameter reduction in one column and cites Han et al. for '6% parameter reduction,' contradicting the earlier statement that pruning reaches 90% in Section 8; and the footnote to Table 3 promises a derivation of recall in Appendix A, but no Appendix A appears in the manuscript.
minor comments (4)
  1. [Abstract] The claim of 'improving generalization capability' is based on a single run; the authors should either temper the language or provide repeated runs with variance estimates.
  2. [§5] The experimental details are underspecified for reproduction: the learning rate, optimizer hyperparameters, data augmentation, input resolution, and hardware are not reported.
  3. [§4] Equation (6) defines G(x), but the text subsequently uses G_s and 'rootstock model' interchangeably; the distinction between the rootstock and the grafted model should be clarified.
  4. [References] References [13] and [15] appear to be the same paper published in two venues; please distinguish or merge them.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline accuracy and size numbers are direct experimental/arithmetic outputs, not predictions forced by the paper's equations or by self-citation.

full rationale

The mathematical framework in Eqs. (1)-(8) defines a combinatorial selection problem but is never used to select grafting points; Section 9 states 'Our current implementation relies on manually selecting layers and features with predetermined grafting points.' The reported 89.97% validation and 90.45% test accuracies are measured after training, and the 88.54% reduction is arithmetic on Table 2's parameter counts. No fitted parameter is later relabeled as a prediction, and no equation is equivalent to the conclusion. The self-citations ([1], [12], [13], [14]) are motivational (rural-AI constraints, prior weed-detection infrastructure, feature-selection time savings) and do not carry the main claim. The paper's main weakness is an internal inconsistency: Section 5 Step 3 says scion layers are 'selected from the rootstock model,' while Step 4 and Eqs. (3)-(6) describe donor features; if donor layers 8-10 are required at inference, the reported 7.38MB footprint omits them. That is a correctness/attribution concern (there is no rootstock-only baseline), not a circular reduction, so it does not increase the circularity score.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The method depends on manually chosen scion layers and hyperparameters, and on the unverified assumptions that donor features transfer beneficially and that the small model's parameter count reflects the deployed system. No code or ablations are provided, so the ledger is dominated by choices rather than derived constraints.

free parameters (4)
  • Selected scion layer indices S = {8, 9, 10} = 8, 9, 10
    Chosen by hand rather than by solving the optimization in Eqs. (7)-(8); the paper gives no search or sensitivity analysis, and these choices determine both the model size and the reported accuracy.
  • Grafting head dense width = 256 units
    Dense layers with 256 units are used in both the donor and scion heads (Section 5); no ablation justifies this capacity.
  • Training epochs = 18
    Training is stopped at 18 epochs, and the paper notes validation was still improving, so the reported results depend on this cutoff.
  • Batch size = 16
    Batch size 16 is used in the rootstock training step; no sensitivity analysis is provided.
assumptions (3)
  • ad hoc to paper Intermediate feature maps from layers 8, 9, and 10 of the VGG16 donor can be pooled, concatenated, and used to improve a smaller rootstock classifier.
    This is the central mechanism of the method, asserted in Section 5 Step 3 and never tested against a rootstock without the scion.
  • domain assumption The grafted model G(x) = M(C({T(f_i(x)) | i in S})) is trainable end-to-end and compatible in shape.
    Section 4 assumes the grafted model is a valid neural network; no compatibility or gradient-flow analysis is given for concatenating donor features with rootstock features.
  • domain assumption Validation accuracy on a single 60/20/20 split is a reliable proxy for deployment performance.
    Section 5.2 uses one split without cross-validation, and Section 6.2 treats test accuracy on this split as conclusive evidence of generalization.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Knowledge Grafting: A Mechanism for Optimizing AI Model Deployment in Resource-Constrained Environments." pith.science (2026). https://pith.science/paper/W6XZJC3Z

@misc{pith2026250719261,
  author       = {Pith},
  title        = {Pith review of: Knowledge Grafting: A Mechanism for Optimizing AI Model Deployment in Resource-Constrained Environments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W6XZJC3Z}},
  note         = {Machine review of arXiv:2507.19261}
}
read the original abstract

The increasing adoption of Artificial Intelligence (AI) has led to larger, more complex models with numerous parameters that require substantial computing power -- resources often unavailable in many real-world application scenarios. Our paper addresses this challenge by introducing knowledge grafting, a novel mechanism that optimizes AI models for resource-constrained environments by transferring selected features (the scion) from a large donor model to a smaller rootstock model. The approach achieves an 88.54% reduction in model size (from 64.39 MB to 7.38 MB), while improving generalization capability of the model. Our new rootstock model achieves 89.97% validation accuracy (vs. donor's 87.47%), maintains lower validation loss (0.2976 vs. 0.5068), and performs exceptionally well on unseen test data with 90.45% accuracy. It addresses the typical size vs performance trade-off, and enables deployment of AI frameworks on resource-constrained devices with enhanced performance. We have tested our approach on an agricultural weed detection scenario, however, it can be extended across various edge computing scenarios, potentially accelerating AI adoption in areas with limited hardware/software support -- by mirroring in a similar manner the horticultural grafting enables productive cultivation in challenging agri-based environments.

Figures

Figures reproduced from arXiv: 2507.19261 by the authors.

Figure 1
Figure 1. A depiction of grafting process in plants. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Grafting process in a neural network [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Training and validation metrics for donor and rootstock models across 18 epochs: (a) [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 20 canonical work pages

  1. [1]

    Adaptive edge-cloud environments for rural ai

    Osama Almurshed, Ashish Kaushal, Nitin Auluck, and Omer F Rana. Adaptive edge-cloud environments for rural ai. InProceedings of the IEEE International Conference on Services Computing (SCC), pages 74–83. IEEE, 2022

    work page 2022

  2. [2]

    Lan- guage models are few-shot learners.arXiv preprint arXiv:2005.14165, 2020

    Tom B Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Lan- guage models are few-shot learners.arXiv preprint arXiv:2005.14165, 2020

    arXiv 2005

  3. [3]

    An analysis of deep neural network models for practical applications.arXiv preprint arXiv:1605.07678, 2016

    Alfredo Canziani, Adam Paszke, and Eugenio Culurciello. An analysis of deep neural network models for practical applications.arXiv preprint arXiv:1605.07678, 2016

    arXiv 2016

  4. [4]

    Evaluating deep learning models for effective weed classification in agricultural images

    Bianca Panacho Ferreira, Pedro Henrique Campos Moreira, Leandro Henrique Fur- tado Pinto Silva, and João Fernando Mari. Evaluating deep learning models for effective weed classification in agricultural images. Revista de Informática Teórica e Aplicada, 32 (1):265–272, 2025

    work page 2025

  5. [5]

    Dynamic network surgery for efficient dnns

    Yiwen Guo, Anbang Yao, and Yurong Chen. Dynamic network surgery for efficient dnns. Advances in Neural Information Processing Systems, 29:1379–1387, 2016

    work page 2016

  6. [6]

    Deep compression: Compressing deep neu- ral networks with pruning, trained quantization and huffman coding

    Song Han, Huizi Mao, and William J Dally. Deep compression: Compressing deep neu- ral networks with pruning, trained quantization and huffman coding. arXiv preprint arXiv:1510.00149, 2015

    arXiv 2015

  7. [7]

    Learning both weights and connec- tions for efficient neural network

    Song Han, Jeff Pool, John Tran, and William Dally. Learning both weights and connec- tions for efficient neural network. InAdvances in Neural Information Processing Systems, volume 28, pages 1135–1143, 2015

    work page 2015

  8. [8]

    Distilling the knowledge in a neural network

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015

    arXiv 2015

Show all 37 references
  1. [9]

    Mobilenets: Efficient convolutional neural networks for mobile vision applications.arXiv preprint arXiv:1704.04861, 2017

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

    2017 arXiv

  2. [10]

    Quantization and training of neu- ral networks for efficient integer-arithmetic-only inference

    Benoit Jacob, Skirmantas Kligys, Bo Chen, Menglong Zhu, Matthew Tang, Andrew Howard, Hartwig Adam, and Dmitry Kalenichenko. Quantization and training of neu- ral networks for efficient integer-arithmetic-only inference. In Proceedings of the IEEE Conference on Computer Vision ...

    2018

  3. [11]

    Spherical linear interpolation and bézier curves.General Scientific Researches, 2(1):13–17, 2014

    Mehdi Jafari and Habib Molaei. Spherical linear interpolation and bézier curves.General Scientific Researches, 2(1):13–17, 2014

    2014

  4. [12]

    An edge-cloud infrastructure for weed detection in precision agriculture

    Ashish Kaushal, Osama Almurshed, Areej Alabbas, Nitin Auluck, and Omer Rana. An edge-cloud infrastructure for weed detection in precision agriculture. In2023 IEEE Intl Conf on Dependable, Autonomic and Secure Computing, Intl Conf on Pervasive Intelli- gence and Computing, Intl...

  5. [13]

    Tosim- iot: Towardsasustainableoptimisationofmachinelearningtasksininternetofthings

    Ashish Kaushal, Osama Almurshed, Asmail Muftah, Nitin Auluck, and Omer Rana. Tosim- iot: Towardsasustainableoptimisationofmachinelearningtasksininternetofthings. IEEE Internet of Things Journal, 2024

    2024

  6. [14]

    Shield: A secure heuristic integrated environment for load distribution in rural-ai.Future Generation Computer Systems, 2024

    Ashish Kaushal, Osama Almurshed, et al. Shield: A secure heuristic integrated environment for load distribution in rural-ai.Future Generation Computer Systems, 2024

    2024

  7. [15]

    Tosim- iot: Towardsasustainableoptimisationofmachinelearningtasksininternetofthings

    Ashish Kaushal, Osama Almurshed, Asmail Muftah, Nitin Auluck, and Omer Rana. Tosim- iot: Towardsasustainableoptimisationofmachinelearningtasksininternetofthings. IEEE Internet of Things Journal, 2025

    2025

  8. [16]

    Do better imagenet models transfer better? In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2661–2671, 2019

    Simon Kornblith, Jonathon Shlens, and Quoc V Le. Do better imagenet models transfer better? In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2661–2671, 2019

    2019

  9. [17]

    Quantifying the carbon emissions of machine learning.arXiv preprint arXiv:1910.09700, 2019

    Alexandre Lacoste, Alexandra Luccioni, Victor Schmidt, and Thomas Dandres. Quantifying the carbon emissions of machine learning.arXiv preprint arXiv:1910.09700, 2019

    1910 arXiv

  10. [18]

    Crossfuse: A novel cross attention mechanism based infrared and visible image fusion approach.Information Fusion, 103:102147, 2024

    Hui Li and Xiao-Jun Wu. Crossfuse: A novel cross attention mechanism based infrared and visible image fusion approach.Information Fusion, 103:102147, 2024

    2024

  11. [19]

    Prun- ing and quantization for deep neural network acceleration: A survey

    Tailin Liang, John Glossner, Lei Wang, Shaobo Shi, and Xiaotong Zhang. Prun- ing and quantization for deep neural network acceleration: A survey. arXiv preprint arXiv:2101.09671, 2021

    2021 arXiv

  12. [20]

    Learning efficient convolutional networks through network slimming

    Zhuang Liu, Jianguo Li, Zhiqiang Shen, Gao Huang, Shoumeng Yan, and Changshui Zhang. Learning efficient convolutional networks through network slimming. InProceedings of the IEEE International Conference on Computer Vision, pages 2736–2744, 2017

    2017

  13. [21]

    Matching the ideal pruning method with knowledge distillation for optimal compression

    Leila Malihi and Gunther Heidemann. Matching the ideal pruning method with knowledge distillation for optimal compression. Applied System Innovation, 7(4):56, 2024. doi: 10. 3390/asi7040056. URL https://www.mdpi.com/2571-5577/7/4/56

    2024

  14. [22]

    Augmented language models: a survey.arXiv preprint arXiv:2302.07842, 2023

    Grégoire Mialon, Roberto Dessì, Maria Lomeli, Christoforos Nalmpantis, Ram Pasunuru, Roberta Raileanu, Baptiste Rozière, Timo Schick, Jane Dwivedi-Yu, Asli Celikyilmaz, et al. Augmented language models: a survey.arXiv preprint arXiv:2302.07842, 2023

    2023 arXiv

  15. [23]

    Konovalov, Bronson Philippa, Peter Ridd, Jake C

    Alex Olsen, Dmitry A. Konovalov, Bronson Philippa, Peter Ridd, Jake C. Wood, Jamie Johns, Wesley Banks, Benjamin Girgenti, Owen Kenny, James Whinney, Brendan Calvert, Mostafa Rahimi Azghadi, and Ronald D. White. DeepWeeds: A Multiclass Weed Species Image Dataset for Deep Learn...

    2019 doi

  16. [24]

    Rural ai: Serverless-powered federated learning for remote applications.IEEE Internet Computing, 27(2):28–34, 2022

    Panos Patros, Ahmed Al-Hafidh, Sayan Mazumdar, Abdullah Alamgir, Mohammad Alhus- ban, Jim Alzahrani, James Frampton, and Albert Y Zomaya. Rural ai: Serverless-powered federated learning for remote applications.IEEE Internet Computing, 27(2):28–34, 2022

    2022

  17. [25]

    Model compression via distillation and quantization

    Antonio Polino, Razvan Pascanu, and Dan Alistarh. Model compression via distillation and quantization. In International Conference on Learning Representations, 2018

    2018

  18. [26]

    Fitnets: Hints for thin deep nets

    Adriana Romero, Nicolas Ballas, Samira Ebrahimi Kahou, Antoine Chassang, Carlo Gatta, and Yoshua Bengio. Fitnets: Hints for thin deep nets. arXiv preprint arXiv:1412.6550, 2014

    2014 arXiv

  19. [27]

    Very deep convolutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014. 17

    2014 arXiv

  20. [28]

    Energy and policy considerations for deep learning in nlp.arXiv preprint arXiv:1906.02243, 2019

    Emma Strubell, Ananya Ganesh, and Andrew McCallum. Energy and policy considerations for deep learning in nlp.arXiv preprint arXiv:1906.02243, 2019

    1906 arXiv

  21. [29]

    Efficientnet: Rethinking model scaling for convolutional neural networks

    Mingxing Tan and Quoc Le. Efficientnet: Rethinking model scaling for convolutional neural networks. International Conference on Machine Learning, pages 6105–6114, 2019

    2019

  22. [30]

    Post-training quantization

    TensorFlow Documentation. Post-training quantization. 2023

    2023

  23. [31]

    Tripp, Jordan Perr-Sauer, Luke Hayne, Monte Lunacek, Ryan N

    Christopher E. Tripp, Jordan Perr-Sauer, Luke Hayne, Monte Lunacek, Ryan N. King, Andrew Glaws, and Ryan M. Camacho. Measuring the energy consumption and efficiency of deep neural networks: An empirical analysis and design recommendations.arXiv preprint arXiv:2403.08151, 2024....

    2024 arXiv

  24. [32]

    Combining multi-objective genetic algorithm and neural network dynamically for the com- plex optimization problems in physics.Scientific Reports, 13(1):1463, 2023

    Peng Wang, Kai Ye, Xian Hao, Wenbo Zhang, Yanwei Ma, Huafu Ouyang, and Tao Xiong. Combining multi-objective genetic algorithm and neural network dynamically for the com- plex optimization problems in physics.Scientific Reports, 13(1):1463, 2023. doi: 10.1038/ s41598-023-27478-...

    2023

  25. [33]

    Routing experts: Learning to route dynamic experts in existing multi-modal large language models

    Qihan Wu, Zonghan Ke, Yiran Zhou, Xiaodi Sun, and Rongrong Ji. Routing experts: Learning to route dynamic experts in existing multi-modal large language models. In The Thirteenth International Conference on Learning Representations (ICLR), 2025. URL https://arxiv.org/abs/2407....

    2025 arXiv

  26. [34]

    Designing energy-efficient convolutional neural networks using energy-aware pruning

    Tien-Ju Yang, Yu-Hsin Chen, and Vivienne Sze. Designing energy-efficient convolutional neural networks using energy-aware pruning. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 5687–5695, 2017

    2017

  27. [35]

    Scalify: Scale propagation for efficient low-precision llm training.arXiv preprint arXiv:2407.17353, 2024

    Zhewei Yang, Yuxin Shen, Yidi Chen, Hanh Dinh-Xuan, Xueqiao Wang, Pengzhan Feng, Ji Lin, Yifan Wang, Xin Liu, Liang Chen, et al. Scalify: Scale propagation for efficient low-precision llm training.arXiv preprint arXiv:2407.17353, 2024

    2024 arXiv

  28. [36]

    Howtransferablearefeaturesin deep neural networks?Advances in Neural Information Processing Systems, 27:3320–3328, 2014

    JasonYosinski, JeffClune, YoshuaBengio, andHodLipson. Howtransferablearefeaturesin deep neural networks?Advances in Neural Information Processing Systems, 27:3320–3328, 2014

    2014

  29. [37]

    Deep neural networks with multi- branch architectures are intrinsically less non-convex

    Hongyang Zhang, Junru Shao, and Ruslan Salakhutdinov. Deep neural networks with multi- branch architectures are intrinsically less non-convex. InProceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics (AISTATS), volume 89 of Proceedi...

    2019

Pith tools

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