Pith. sign in

REVIEW 4 major objections 6 minor 64 references

Towards Sustainable Artificial Intelligence: A Comprehensive Review and Comparative Analysis of Deep Learning Models' Carbon Footprint

T0 review · 4 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 In CPU benchmarks, training dwarfs other phases in AI models' carbon cost

desk verdict A competent but narrow Green AI survey with a small CPU benchmark; the training-dominates finding is solid in context, but the complexity-vs-accuracy claim is undercut by single runs and missing training details. read the letter →

arxiv 2608.09998 v1 pith:PONVHIW5 submitted 2026-08-07 cs.AI cs.CYcs.LGcs.SE

Samar Garrab , Sarra Boughriou , Manel BenSassi This is my paper · ORCID
classification cs.AIcs.CYcs.LGcs.SE
keywords GreenArtificialIntelligenceDeepLearningCarbonfootprintassessmenttoolsSustainabledevelopmentModeloptimizationEmissionsperAccuracyPointCodeenergyefficiency
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that the carbon footprint of deep learning models should be a first-class selection criterion, alongside accuracy. It backs the argument with a controlled CPU-based benchmark of six convolutional architectures on the CIFAR-10 image set, measured with CodeCarbon at a fixed regional carbon intensity. The results show that the training phase produces the overwhelming share of emissions, and that heavier architectures do not buy proportionally better accuracy. The paper introduces Emissions per Accuracy Point (EAP), grams of CO2e per accuracy point, as a way to compare models on sustainability, and ranks the six models by it. If the finding holds, model choice and optimization effort should be guided by carbon efficiency, not accuracy alone.

What carries the argument

The carrying mechanism is a pair of ratios. CodeCarbon estimates emissions with $C = E \times CI$, where $E$ is energy consumed and $CI$ is regional carbon intensity (39.49 gCO2e/kWh for Ontario, Canada); the paper uses this to attribute emissions per phase. The comparative metric is EAP, $EAP_i = C_i / A_i$, with a min-max normalized version $EAP^{Norm}_i = (EAP_i - \min(EAP)) / (\max(EAP) - \min(EAP))$. EAP does the argumentative work: it turns raw emissions into a performance-adjusted quantity, so a high-accuracy model with high emissions (ResNet) can be ranked below a lighter model (U-Net) once carbon cost per accuracy point is considered.

What would settle it

Run each of the six models multiple times (for example, ten runs) under the same CPU conditions and compare run-to-run variation in emissions and accuracy. If the spread within a model matches or exceeds the differences between models, such as EfficientNet's 4.67 g versus Simple CNN's 6.02 g, the EAP ranking is not stable. A second decisive test is repeating the benchmark on a GPU or cloud instance; if training no longer dominates the lifecycle, the paper's central claim is specific to CPU settings.

Watch

Extended reading notes

Core claim

On the authors' CPU-based experimental setup, measured carbon emissions ranged from 1.76 gCO2e for U-Net to 15.64 gCO2e for VGG19 across the full pipeline, with training contributing orders of magnitude more than preprocessing and testing. Accuracy did not track emissions: ResNet reached the highest accuracy (90%) at 12.34 g, while VGG19 combined the highest emissions with the lowest accuracy (78%), and U-Net obtained 84% accuracy for less than one-seventh of VGG19's emissions. The paper's central claim is that training is the primary source of emissions in this setting and that increased architectural complexity does not systematically translate into proportional accuracy gains. It formalizes the accuracy–emissions trade-off through EAP, defined as total emissions divided by accuracy, with lower values indicating better carbon efficiency; normalized EAP scores place U-Net at 0.00 and VGG19 at 1.00.

Load-bearing premise

The benchmark's conclusions rest on the assumption that one execution of each model, on one machine and one dataset, gives a representative picture of both accuracy and emissions; the paper itself notes that measurements came from single experimental runs.

Editorial extensions

If this is right

  • Optimization effort in CPU-bound training should concentrate on the training phase, since preprocessing and testing contribute emissions on the order of $10^{-3}$ to $10^{-1}$ g.
  • Model selection should report EAP alongside accuracy; a lighter architecture can deliver competitive accuracy at a fraction of the carbon cost.
  • Architectural complexity is not a reliable proxy for predictive gain: the highest-emission model in the benchmark had the lowest accuracy.
  • Carbon-aware benchmarking, where emissions are reported with performance metrics, becomes a practical complement to accuracy-only reporting.
  • The quantitative rankings are directly applicable to small-to-moderate scale CPU training; the authors caution that GPU, cloud, and distributed settings may differ qualitatively.

Reading between the lines

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

  • If repeated runs confirm the single-run measurements, the EAP ranking is robust enough to inform model selection in resource-constrained settings; a natural next step is computing confidence intervals from several runs per model.
  • The pattern may shift at deployment scale: literature cited by the paper reports inference as roughly 65% of lifecycle emissions in production, so a CPU-training benchmark likely underestimates the relative importance of inference for always-on services.
  • EAP generalizes beyond the six architectures tested; applying it to GPU or cloud benchmarks would test whether training-dominance is a property of CPUs or of deep learning generally.
  • The measurement pipeline of CodeCarbon plus EAP could be extended to pruning, quantization, and mixed-precision training, with the testable prediction that these techniques reduce EAP mainly by cutting training-phase energy.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper combines a PRISMA-based systematic literature review of Green AI / Green DL and a comparative analysis of seven carbon-footprint measurement tools with an empirical study that measures the carbon emissions of six CNN architectures (Simple CNN, U-Net, ResNet, VGG16, VGG19, EfficientNet) on CIFAR-10 using CodeCarbon on a CPU. It reports total and per-phase emissions (preprocessing, training, testing), defines an Emissions per Accuracy Point (EAP) metric, and concludes that the training phase dominates emissions and that increased architectural complexity does not proportionally improve accuracy, motivating carbon-aware model selection.

Significance. If the quantitative claims hold, the paper makes a useful, reproducible contribution: the experimental setup, CodeCarbon measurements, per-phase breakdown, and EAP rankings provide a concrete benchmark for CPU-based green DL evaluation, and the literature review with tool comparison is a helpful synthesis. The central qualitative finding that training dominates per-run emissions in this setup is strongly supported by orders-of-magnitude differences, and the paper is unusually transparent in its code/data availability and in its stated limitations. However, the accuracy-complexity claim and the EAP ranking are more fragile because they rest on single runs with unreported training hyperparameters, and several terminological and unit errors weaken the presentation.

major comments (4)
  1. [Section 4.2, Section 7 (Internal Validity), Table 7] The entire quantitative comparison rests on a single execution per model, and the training protocol is not reported: epochs, batch size, optimizer, learning rate, early stopping, and validation-based model selection are all absent. Because training emissions scale with the number of optimization steps and accuracy depends strongly on the optimization schedule, the central claim that 'increased architectural complexity does not systematically translate into proportional accuracy gains' is not distinguished from the possibility that the larger models (e.g., VGG19 at 78%) were undertrained or affected by seed variability. Please report the full hyperparameters, repeat the runs, and provide variance estimates or at least a sensitivity discussion; without this, the EAP ranking in Table 7 and the associated model-selection advice are not load-bearing.
  2. [Section 5, Eq. (1), Table 7] EAP is defined as C_i / A_i, but A_i is reported as a fraction (e.g., 0.84 for U-Net), so the values in Table 7 (e.g., 2.10) are grams of CO2e per unit accuracy, not 'grams per accuracy point.' If 'accuracy point' means a percentage point, the denominator should be 84, giving 0.02095 g/point. The normalized EAP ranking is invariant to this uniform scaling, but the units and the text's interpretation of EAP as 'grams of CO2e emitted per accuracy point' are incorrect and should be corrected.
  3. [Abstract, Section 4.3.2, Section 4.1] The empirical task is repeatedly described as 'multi-label classification,' but CIFAR-10 with one-hot categorical encoding is a single-label, multi-class classification problem. The paper itself states that each image has a single class and uses 'to_categorical' for one-hot encoding. The terminology should be corrected throughout, or the study should actually include a multi-label setting; as written, the stated experimental scope does not match the data used.
  4. [Section 5, Section 8, Abstract] The conclusion that 'the training phase is the primary source of emissions' is presented as a lifecycle finding, but the measurements cover only one training run and one test pass with no deployment or inference-lifetime analysis. The paper's own literature review (Section 2.3.1) cites Schwartz et al. and Wu et al., who report that inference can dominate in deployed systems (Wu et al. report ~65% inference versus 35% training). The claim should be explicitly restricted to 'per-run emissions in this CPU-based experimental setting' in the abstract and conclusion, not stated as a general lifecycle result.
minor comments (6)
  1. [Section 2.1, Fig. 2] The PRISMA flow diagram numbers are internally consistent, but the text could state the final numbers of included studies from database search and snowballing separately; currently they must be inferred from the diagram.
  2. [Section 3.2, Table 5] The comparison criteria and table are useful, but the abbreviations 'C1'–'C6' are reintroduced in the text without being explicitly tied to the table rows for readers skimming the narrative.
  3. [References [59], [60], [61]] EfficientNet is cited to a rice-leaf-disease paper rather than the original EfficientNet paper (Tan and Le, 2019), and VGG16/VGG19 are cited to secondary sources rather than the original Simonyan and Zisserman work; please cite the primary architecture papers.
  4. [Section 4.3.2] The data preparation section reports that blurry and dark images were removed but gives no counts or thresholds; please state how many images were removed and how 'blurry' and 'dark' were defined.
  5. [Section 4.2] The hardware description should specify whether CPU frequency scaling, turbo boost, and background processes were controlled beyond 'unnecessary background processes were minimized,' since these can materially affect energy measurements on a local machine.
  6. [Throughout] There are several typographical artifacts in the tables and captions, such as 'T able 1,' 'T racker,' and 'F unctionality'; these should be cleaned up in the final version.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical claims are direct measurements with CodeCarbon, and EAP is an explicit ratio, not a fitted or self-referential quantity.

full rationale

The paper's central claims are empirical rather than derivational. Emissions are measured with an external tool (CodeCarbon) using C=E×CI, and the per-phase breakdown in Fig. 5 is a direct measurement. The EAP metric in Eq. (1) is defined as EAP_i = C_i / A_i, a simple ratio of two independently measured quantities; it is not fitted to the conclusions, and the 'training phase dominates' conclusion does not follow from the EAP definition. The accuracy-vs-complexity observation is a reported pattern over six executed models, not a consequence of the chosen metric. The single-run design and missing training details are genuine threats to validity and generalizability, but they concern measurement reliability, not circularity. No load-bearing self-citation exists: EAP is attributed to prior work [3,4,64] only as provenance for a standard ratio, and no uniqueness theorem or fitted parameter forces the ranking. Therefore the circularity burden is zero.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities. Its central claim depends on the measurement tool (CodeCarbon), the regional carbon intensity, single-run execution, and the choice of architectures and dataset, all of which are domain assumptions. The only hidden degree of freedom is the unreported training configuration.

free parameters (1)
  • Training hyperparameters (epochs, batch size, optimizer, learning rate) = Not reported
    The empirical comparison depends on training configurations that are never disclosed. Different choices would change both accuracy and energy consumption, so the cross-model ranking is underdetermined.
assumptions (6)
  • domain assumption CodeCarbon accurately estimates energy consumption and carbon emissions for CPU workloads.
    Section 4.2 uses CodeCarbon as the sole measurement tool; if its estimates are biased, all reported emissions and the EAP ranking change.
  • domain assumption The Ontario carbon intensity of 39.49 gCO2e/kWh is representative of the execution environment.
    Section 4.2 sets CI from CodeCarbon for Ontario; emissions scale linearly with CI, so results would shift under a different grid.
  • domain assumption A single run per model yields representative measurements.
    Section 4.2 states each model was executed once; Section 7 acknowledges this as a threat to internal validity.
  • domain assumption The six selected architectures fairly represent the complexity spectrum of deep learning models.
    Section 4.1 selects a CNN, U-Net, ResNet, EfficientNet, VGG16, and VGG19; conclusions about complexity and accuracy depend on this choice.
  • domain assumption CIFAR-10 test accuracy is a valid measure of predictive performance for comparing models.
    Section 4.3 uses CIFAR-10 and test accuracy as the performance metric; other datasets or metrics could alter the accuracy-emissions trade-off.
  • domain assumption The SLR inclusion/exclusion criteria capture the relevant Green AI literature.
    Section 2.1 applies PRISMA and manual screening; the 49 included studies shape the review conclusions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Sustainable Artificial Intelligence: A Comprehensive Review and Comparative Analysis of Deep Learning Models' Carbon Footprint." pith.science (2026). https://pith.science/paper/PONVHIW5

@misc{pith2026260809998,
  author       = {Pith},
  title        = {Pith review of: Towards Sustainable Artificial Intelligence: A Comprehensive Review and Comparative Analysis of Deep Learning Models' Carbon Footprint},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PONVHIW5}},
  note         = {Machine review of arXiv:2608.09998}
}
read the original abstract

Artificial Intelligence (AI) and Machine Learning (ML) have become powerful tools for supporting and automating complex human tasks. Despite their benefits, growing attention has been directed toward their environmental implications, primarily due to their high energy demands and associated carbon emissions. This concern is particularly relevant in light of the increasing deployment of large-scale models, especially Deep Learning (DL) architectures, which provide advanced predictive capabilities but require substantial computational resources. This paper presents a systematic review of research on Green AI, Green DL, and optimization techniques aimed at reducing the environmental impact of AI models. In addition, we examine and compare several carbon measurement tools for estimating emissions generated by AI algorithms. To complement the review, we conducted an empirical evaluation using a CPU-based experimental setup, in which six DL models were implemented for a multi-label classification task. The objective was to quantify and compare their overall carbon emissions and to determine which stages of the DL lifecycle contribute most significantly to the total footprint. The results show that the training phase is the primary source of emissions. Moreover, the findings reveal that increased architectural complexity does not systematically translate into proportional accuracy gains, highlighting the importance of carefully balancing predictive performance and environmental cost. These results reinforce the need to integrate sustainability considerations into model selection and AI system design.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 23 canonical work pages

  1. [1]

    Wiley Inter- disciplinary Reviews: Data Mining and Knowledge Discovery13(4), 1507 (2023) https://doi.org/10.1002/widm.1507

    Verdecchia, R., Sallou, J., Cruz, L.: A systematic review of green ai. Wiley Inter- disciplinary Reviews: Data Mining and Knowledge Discovery13(4), 1507 (2023) https://doi.org/10.1002/widm.1507

    doi:10.1002/widm.1507 2023

  2. [2]

    ACM Computing Surveys (CSUR) 55(2), 1–96 (2022) https://doi.org/10.48550/arXiv.1906.05433

    Rolnick, D., Donti, P.L., Kaack, L.H., Kochanski, K., Lacoste, A., Sankaran, K., Ross, A.S., Milojevic-Dupont, N., Jaques, N., Waldman-Brown, A.,et al.: Tack- ling climate change with machine learning. ACM Computing Surveys (CSUR) 55(2), 1–96 (2022) https://doi.org/10.48550/arXiv.1906.05433

    arXiv doi:10.48550/arxiv.1906.05433 2022

  3. [3]

    In: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp

    Strubell, E., Ganesh, A., McCallum, A.: Energy and policy considerations for deep learning in nlp. In: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 3645–3650 (2019). https://doi.org/10.48550/ arXiv.1906.02243

    arXiv doi:10.48550/arxiv.1906.02243 2019

  4. [4]

    Journal of Machine Learning Research21(248), 1–43 (2020) https://doi.org/10

    Henderson, P., Hu, J., Romoff, J., Brunskill, E., Jurafsky, D., Pineau, J.: Towards the systematic reporting of the energy and carbon footprints of machine learning. Journal of Machine Learning Research21(248), 1–43 (2020) https://doi.org/10. 48550/arXiv.2002.05651

    arXiv doi:10.48550/arxiv.2002.05651 2020

  5. [5]

    arXiv preprint (2022) https://doi

    Pachot, A., Patissier, C.: Towards sustainable artificial intelligence: an overview of environmental protection uses and issues. arXiv preprint (2022) https://doi. org/10.48550/arXiv.2212.11738

    work page doi:10.48550/arxiv.2212.11738 2022

  6. [6]

    Nature Publishing Group UK London

    Dhar, P.: The carbon impact of artificial intelligence. Nature Publishing Group UK London. (2020). https://doi.org/10.1038/s42256-020-0219-9

    doi:10.1038/s42256-020-0219-9 2020

  7. [7]

    Communications of the ACM63(12), 54–63 (2020) https://doi.org/10.1145/3381831

    Schwartz, R., Dodge, J., Smith, N.A., Etzioni, O.: Green ai. Communications of the ACM63(12), 54–63 (2020) https://doi.org/10.1145/3381831

    doi:10.1145/3381831 2020

  8. [8]

    arXiv preprint (2021) https://doi.org/10.48550/arXiv.2111.05193

    Xu, J., Zhou, W., Fu, Z., Zhou, H., Li, L.: A survey on green deep learning. arXiv preprint (2021) https://doi.org/10.48550/arXiv.2111.05193

    arXiv doi:10.48550/arxiv.2111.05193 2021

Show all 64 references
  1. [9]

    bmj372 (2021) https://doi.org/10.1136/bmj.n71

    Page, M.J., McKenzie, J.E., Bossuyt, P.M., Boutron, I., Hoffmann, T.C., Mulrow, C.D., Shamseer, L., Tetzlaff, J.M., Akl, E.A., Brennan, S.E., et al.: The prisma 2020 statement: an updated guideline for reporting systematic reviews. bmj372 (2021) https://doi.org/10.1136/bmj.n71

    2021 doi

  2. [10]

    In: Proceedings of the 18th International Con- ference on Evaluation and Assessment in Software Engineering, pp

    Wohlin, C.: Guidelines for snowballing in systematic literature studies and a replication in software engineering. In: Proceedings of the 18th International Con- ference on Evaluation and Assessment in Software Engineering, pp. 1–10 (2014). https://doi.org/10.1145/2601248.2601268

    2014

  3. [11]

    AI and Ethics1(3), 213–218 (2021) https://doi.org/10.1007/ s43681-021-00043-6 24

    Van Wynsberghe, A.: Sustainable ai: Ai for sustainability and the sustain- ability of ai. AI and Ethics1(3), 213–218 (2021) https://doi.org/10.1007/ s43681-021-00043-6 24

    2021

  4. [12]

    AMCIS 2022 Proceedings (2022)

    Natarajan, H.K., Paula, D., Dremel, C., Uebernickel, F.: A theoretical review on ai affordances for sustainability. AMCIS 2022 Proceedings (2022)

    2022

  5. [13]

    IEEE Access12, 23989–24013 (2024) https://doi.org/10.1109/ ACCESS.2024.3360705

    Barbierato, E., Gatti, A.: Toward green ai: A methodological survey of the sci- entific literature. IEEE Access12, 23989–24013 (2024) https://doi.org/10.1109/ ACCESS.2024.3360705

    2024

  6. [14]

    https://datacenters.google/ operating-sustainably

    Google: Operating sustainably. https://datacenters.google/ operating-sustainably. Google Data Centers (2024)

    2024

  7. [15]

    Reuters (2024)

    Groom, N.: Google partners with Nevada utility for geothermal to power data centers. Reuters (2024). https://www.reuters.com/business/energy/ google-partners-with-nevada-utility-geothermal-power-data-centers-2024-06-13

    2024

  8. [16]

    DataCenterDynamics (2024)

    DCD News: Microsoft signs solar PPAs totaling 475MW with AES. DataCenterDynamics (2024). https://www.datacenterdynamics.com/en/news/ microsoft-signs-solar-ppas-totaling-475mw-with-aes

    2024

  9. [17]

    In: ECAI 2024, pp

    Trzcinski, T., Twardowski, B., Zieli´ nski, B., Adamczewski, K., W´ ojcik, B.: Zero- waste machine learning. In: ECAI 2024, pp. 43–49. IOS Press, Amsterdam (2024). https://doi.org/10.3233/F AIA240466

    2024 doi

  10. [18]

    Emerging Science Journal8(4), 1602–1621 (2024) https://doi.org/10.28991/ESJ-2024-08-04-021

    Acosta-Vargas, P., Salvador-Acosta, B., Novillo-Villegas, S., Sarantis, D., Salvador-Ullauri, L.: Generative artificial intelligence and web accessibility: Towards an inclusive and sustainable future. Emerging Science Journal8(4), 1602–1621 (2024) https://doi.org/10.28991/ESJ-...

    2024 doi

  11. [19]

    HighTech and Innovation Journal 5(4), 1085–1100 (2024) https://doi.org/10.28991/HIJ-2024-05-04-015

    Sookhom, A., Krivart, K., Jansiri, N., Sookhom, A., Chanasit, J., Pathak, S., Mallongi, A.: The new way of tourism in green economy style for sustainable community development and empowerment. HighTech and Innovation Journal 5(4), 1085–1100 (2024) https://doi.org/10.28991/HIJ-...

    2024 doi

  12. [20]

    Emerging Science Journal8(3), 933–947 (2024) https://doi.org/10.28991/ ESJ-2024-08-03-08

    Shaari, M.S., Sulong, A., Ridzuan, A.R., Esquivias, M.A., Lau, E.: The car- bon conundrum: exploring co2 emissions, public debt, and environmental pol- icy. Emerging Science Journal8(3), 933–947 (2024) https://doi.org/10.28991/ ESJ-2024-08-03-08

    2024

  13. [21]

    Proceedings of machine learning and systems4, 795–813 (2022) https://doi.org/10.48550/arXiv.2111.00364

    Wu, C.-J., Raghavendra, R., Gupta, U., Acun, B., Ardalani, N., Maeng, K., Chang, G., Aga, F., Huang, J., Bai, C.,et al.: Sustainable ai: Environmental implications, challenges and opportunities. Proceedings of machine learning and systems4, 795–813 (2022) https://doi.org/10.48...

    2022 doi arXiv

  14. [22]

    arXiv preprint (2019) https://doi.org/10.48550/arXiv

    Lacoste, A., Luccioni, A., Schmidt, V., Dandres, T.: Quantifying the carbon emis- sions of machine learning. arXiv preprint (2019) https://doi.org/10.48550/arXiv. 1910.09700

    2019 doi arXiv

  15. [23]

    arXiv preprint (2023) https://doi.org/10.48550/arXiv.2311.00447

    Zhou, Y., Lin, X., Zhang, X., Wang, M., Jiang, G., Lu, H., Wu, Y., Zhang, K., 25 Yang, Z., Wang, K.,et al.: On the opportunities of green computing: A survey. arXiv preprint (2023) https://doi.org/10.48550/arXiv.2311.00447

    2023 doi arXiv

  16. [24]

    Advanced science8(12), 2100707 (2021) https: //doi.org/10.48550/arXiv.2007.07610

    Lannelongue, L., Grealey, J., Inouye, M.: Green algorithms: quantifying the car- bon footprint of computation. Advanced science8(12), 2100707 (2021) https: //doi.org/10.48550/arXiv.2007.07610

    2021 doi arXiv

  17. [25]

    The VLDB Journal32(4), 791–813 (2023) https://doi.org/10.48550/arXiv.2112.06409

    Whang, S.E., Roh, Y., Song, H., Lee, J.-G.: Data collection and quality challenges in deep learning: A data-centric ai perspective. The VLDB Journal32(4), 791–813 (2023) https://doi.org/10.48550/arXiv.2112.06409

    2023 doi arXiv

  18. [26]

    Mach Learn Tech Rep1(1), 1–6 (2014) https://doi.org/10.13140/RG.2.2.28948.04489

    Ali, P.J.M., Faraj, R.H., Koya, E., Ali, P.J.M., Faraj, R.H.: Data normalization and standardization: a technical report. Mach Learn Tech Rep1(1), 1–6 (2014) https://doi.org/10.13140/RG.2.2.28948.04489

    2014

  19. [27]

    Tall´ on-Ballesteros, A.: The impact of data normalization on the accuracy of machine learning algorithms: A comparative analysis

    Cabello-Solorzano, K., Araujo, I., Pe˜ na, M., Correia, L., J. Tall´ on-Ballesteros, A.: The impact of data normalization on the accuracy of machine learning algorithms: A comparative analysis. In: International Conference on Soft Computing Models in Industrial and Environment...

    2023 doi

  20. [28]

    Neuroinformatics12(2), 229–244 (2014) https://doi.org/10.1007/ s12021-013-9204-3

    Mwangi, B., Tian, T.S., Soares, J.C.: A review of feature reduction techniques in neuroimaging. Neuroinformatics12(2), 229–244 (2014) https://doi.org/10.1007/ s12021-013-9204-3

    2014

  21. [29]

    Journal of Machine Learning Research10(66-71) (2009)

    Van Der Maaten, L., Postma, E., Herik, J.: Dimensionality reduction: a compar- ative review. Journal of Machine Learning Research10(66-71) (2009)

    2009

  22. [30]

    In: International Conference on Artificial Neu- ral Networks, pp

    Hern´ andez-Garc ´ ıa, A., K¨ onig, P.: Further advantages of data augmentation on convolutional neural networks. In: International Conference on Artificial Neu- ral Networks, pp. 95–103 (2018). https://doi.org/10.48550/arXiv.1906.11052 . Springer

    2018 doi arXiv

  23. [31]

    IEEE Transactions on Knowledge and Data Engineering (2025) https://doi.org/10.1109/TKDE.2025

    Wang, Z., Wang, P., Liu, K., Wang, P., Fu, Y., Lu, C.-T., Aggarwal, C.C., Pei, J., Zhou, Y.: A comprehensive survey on data augmentation. IEEE Transactions on Knowledge and Data Engineering (2025) https://doi.org/10.1109/TKDE.2025. 3622600

    2025 doi

  24. [32]

    arXiv preprint (2019) https://doi.org/10.48550/arXiv.1901.09321

    Zhang, H., Dauphin, Y.N., Ma, T.: Fixup initialization: Residual learning without normalization. arXiv preprint (2019) https://doi.org/10.48550/arXiv.1901.09321

    2019 doi arXiv

  25. [33]

    arXiv preprint (2015) https: //doi.org/10.48550/arXiv.1511.06422

    Mishkin, D., Matas, J.: All you need is a good init. arXiv preprint (2015) https: //doi.org/10.48550/arXiv.1511.06422

    2015 doi arXiv

  26. [34]

    arXiv preprint (2015) https://doi.org/10.48550/arXiv

    Ioffe, S.: Batch normalization: Accelerating deep network training by reducing internal covariate shift. arXiv preprint (2015) https://doi.org/10.48550/arXiv. 1502.03167 26

    2015 doi arXiv

  27. [35]

    arXiv preprint (2016) https://doi.org/10.48550/arXiv.1607.06450

    Ba, J.L., Kiros, J.R., Hinton, G.E.: Layer normalization. arXiv preprint (2016) https://doi.org/10.48550/arXiv.1607.06450

    2016 doi arXiv

  28. [36]

    In: Proceedings of the European Conference on Computer Vision (ECCV), pp

    Wu, Y., He, K.: Group normalization. In: Proceedings of the European Conference on Computer Vision (ECCV), pp. 3–19 (2018). https://doi.org/10.48550/arXiv. 1803.08494

    2018 doi arXiv

  29. [37]

    In: International Conference on Machine Learning, pp

    Belilovsky, E., Eickenberg, M., Oyallon, E.: Greedy layerwise learning can scale to imagenet. In: International Conference on Machine Learning, pp. 583–593 (2019). https://doi.org/10.48550/arXiv.1812.11446 . PMLR

    2019 doi arXiv

  30. [38]

    https://openreview.net/forum? id=ryZ8sz-Ab

    Yu, K., Liu, Y., Schwing, A.G., Peng, J.: Fast and Accurate Text Classification: Skimming, Rereading and Early Stopping (2018). https://openreview.net/forum? id=ryZ8sz-Ab

    2018

  31. [39]

    Advances in neural information processing systems2(1989)

    LeCun, Y., Denker, J., Solla, S.: Optimal brain damage. Advances in neural information processing systems2(1989)

    1989

  32. [40]

    ACM Computing Surveys55(12), 1–37 (2023)

    Menghani, G.: Efficient deep learning: A survey on making deep learning models smaller, faster, and better. ACM Computing Surveys55(12), 1–37 (2023)

    2023

  33. [41]

    arXiv preprint (2022) https: //doi.org/10.48550/arXiv.2207.00112

    Hsu, Y.-C., Hua, T., Chang, S., Lou, Q., Shen, Y., Jin, H.: Language model compression with weighted low-rank factorization. arXiv preprint (2022) https: //doi.org/10.48550/arXiv.2207.00112

    2022 doi arXiv

  34. [42]

    arXiv preprint (2018) https://doi.org/10.48550/arXiv.1806

    Krishnamoorthi, R.: Quantizing deep convolutional networks for efficient infer- ence: A whitepaper. arXiv preprint (2018) https://doi.org/10.48550/arXiv.1806. 08342

    2018 doi

  35. [43]

    arXiv preprint, 1–9 (2015) https://doi.org/10.48550/arXiv.1503.02531

    Hinton, G.E., Vinyals, O., Dean, J.: Distilling the knowledge in a neural network. arXiv preprint, 1–9 (2015) https://doi.org/10.48550/arXiv.1503.02531

    2015 doi arXiv

  36. [44]

    Ba, J., Caruana, R.: Do deep nets really need to be deep? Advances in neural information processing systems27(2014) https://doi.org/10.48550/arXiv.1312. 6184

    2014 doi

  37. [45]

    nature521(7553), 436–444 (2015) https://doi.org/10.1038/nature14539

    LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. nature521(7553), 436–444 (2015) https://doi.org/10.1038/nature14539

    2015 doi

  38. [46]

    Li, D., Chen, X., Becchi, M., Zong, Z.: Evaluating the energy efficiency of deep convolutional neural networks on cpus and gpus. In: 2016 IEEE Inter- national Conferences on Big Data and Cloud Computing (BDCloud), Social Computing and Networking (SocialCom), Sustainable Comput...

    2016 doi

  39. [47]

    In: 2022 International Conference on ICT for Sustainability (ICT4S), pp

    Guti´ errez, M., Moraga, M.´A., Garc ´ ıa, F.: Analysing the energy impact of differ- ent optimisations for machine learning models. In: 2022 International Conference on ICT for Sustainability (ICT4S), pp. 46–52 (2022). https://doi.org/10.1109/ ICT4S55073.2022.00016 . IEEE

    2022

  40. [48]

    In: 2023 IEEE 20th International Conference on Software Architecture Companion (ICSA-C), pp

    Omar, R.: Ai and energy efficiency. In: 2023 IEEE 20th International Conference on Software Architecture Companion (ICSA-C), pp. 141–144 (2023). https://doi. org/10.1109/ICSA-C57050.2023.00040 . IEEE

    2023

  41. [49]

    In: Proceedings of the Second Workshop on Simple and Efficient Natural Language Processing, pp

    Bannour, N., Ghannay, S., N´ ev´ eol, A., Ligozat, A.-L.: Evaluating the carbon footprint of nlp methods: a survey and analysis of existing tools. In: Proceedings of the Second Workshop on Simple and Efficient Natural Language Processing, pp. 11–21 (2021). https://doi.org/10.1...

    2021 doi

  42. [50]

    In: Doklady Mathematics, vol

    Budennyy, S.A., Lazarev, V.D., Zakharenko, N.N., Korovin, A.N., Plosskaya, O., Dimitrov, D.V., Akhripkin, V., Pavlov, I., Oseledets, I.V., Barsola, I.S.,et al.: Eco2ai: carbon emissions tracking of machine learning models as the first step towards sustainable ai. In: Doklady M...

    2022 doi

  43. [51]

    arXiv preprint (2019) https://doi.org/10.48550/arXiv.1911.08354

    Lottick, K., Susai, S., Friedler, S.A., Wilson, J.P.: Energy usage reports: Envi- ronmental awareness as part of algorithmic accountability. arXiv preprint (2019) https://doi.org/10.48550/arXiv.1911.08354

    2019 doi arXiv

  44. [52]

    arXiv preprint (2020) https://doi.org/10.48550/arXiv.2007.03051

    Anthony, L.F.W., Kanding, B., Selvan, R.: Carbontracker: Tracking and predict- ing the carbon footprint of training deep learning models. arXiv preprint (2020) https://doi.org/10.48550/arXiv.2007.03051

    2020 doi arXiv

  45. [53]

    Accessed: 2024-06-26 (2023)

    ML CO2: CodeCarbon: Track and reduce your carbon emissions in machine learning. Accessed: 2024-06-26 (2023). https://github.com/mlco2/codecarbon

    2023

  46. [54]

    Semester Project, EPFL (2020)

    Tr´ ebaol, T.: Cumulator—a tool to quantify and report the carbon footprint of machine learning computations and communication in academia and healthcare. Semester Project, EPFL (2020)

    2020

  47. [55]

    In: 2017 IEEE 2nd International Conference on Big Data Analysis (ICBDA), pp

    Guo, T., Dong, J., Li, H., Gao, Y.: Simple convolutional neural network on image classification. In: 2017 IEEE 2nd International Conference on Big Data Analysis (ICBDA), pp. 721–724 (2017). https://doi.org/10.1109/ICBDA.2017.8078730 . IEEE

    2017

  48. [56]

    IEEE Access11, 51199–51213 (2023) https://doi.org/10.1109/ACCESS.2023.3280191

    El-Bana, S., Al-Kabbany, A., Elragal, H.M., El-Khamy, S.: Evaluating the poten- tial of wavelet pooling on improving the data efficiency of light-weight cnns. IEEE Access11, 51199–51213 (2023) https://doi.org/10.1109/ACCESS.2023.3280191

    2023

  49. [57]

    In: International Conference on Medical Image Computing and Computer-assisted Intervention, pp

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomedical image segmentation. In: International Conference on Medical Image Computing and Computer-assisted Intervention, pp. 234–241 (2015). https://doi. 28 org/10.1007/978-3-319-24574-4 28 . Springer

    2015 doi

  50. [58]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recog- nition, pp

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recog- nition, pp. 770–778 (2016). https://openaccess.thecvf.com/content cvpr 2016/ html/He Deep Residual Learning CVPR 2016...

    2016

  51. [59]

    arXiv preprint (2024) https://doi.org/10.48550/arXiv.2408.01752

    Saddami, K., Nurdin, Y., Zahramita, M., Safiruz, M.S.: Advancing green ai: Efficient and accurate lightweight cnns for rice leaf disease identification. arXiv preprint (2024) https://doi.org/10.48550/arXiv.2408.01752

    2024 doi arXiv

  52. [60]

    In: 2018 IEEE 8th Annual Computing and Communication Workshop and Conference (CCWC), pp

    Qassim, H., Verma, A., Feinzimer, D.: Compressed residual-vgg16 cnn model for big data places image recognition. In: 2018 IEEE 8th Annual Computing and Communication Workshop and Conference (CCWC), pp. 169–175 (2018). https://doi.org/10.1109/CCWC.2018.8301729 . IEEE

    2018

  53. [61]

    In: 2021 Interna- tional Conference on Disruptive Technologies for Multi-disciplinary Research and Applications (CENTCON), vol

    Mascarenhas, S., Agarwal, M.: A comparison between vgg16, vgg19 and resnet50 architecture frameworks for image classification. In: 2021 Interna- tional Conference on Disruptive Technologies for Multi-disciplinary Research and Applications (CENTCON), vol. 1, pp. 96–99 (2021). h...

    2021

  54. [62]

    IEEE transactions on pattern analysis and machine intelligence30(11), 1958–1970 (2008) https://doi.org/10

    Torralba, A., Fergus, R., Freeman, W.T.: 80 million tiny images: A large data set for nonparametric object and scene recognition. IEEE transactions on pattern analysis and machine intelligence30(11), 1958–1970 (2008) https://doi.org/10. 1109/TPAMI.2008.128

    2008

  55. [63]

    University of Toronto18268744(2009)

    Krizhevsky, A., Hinton, G., et al.: Learning multiple layers of features from tiny images. University of Toronto18268744(2009)

    2009

  56. [64]

    arXiv preprint arXiv:2104.10350 (2021) https://doi.org/10.48550/arXiv

    Patterson, D., Gonzalez, J., Le, Q., Liang, C., Munguia, L.-M., Rothchild, D., So, D., Texier, M., Dean, J.: Carbon emissions and large neural network training. arXiv preprint arXiv:2104.10350 (2021) https://doi.org/10.48550/arXiv. 2104.10350 29

    2021 doi arXiv

Pith tools

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