Pith. sign in

REVIEW 3 major objections 5 minor 41 references

GreenMachine: Automatic Design of Zero-Cost Proxies for Energy-Efficient NAS

T0 review · 3 major / 5 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 GreenMachine evolves zero-cost proxies that rank untrained neural networks more accurately than existing training-free methods on NATS-Bench, reaching a Kendall correlation of 0.89 on CIFAR-10.

desk verdict Evolved zero-cost proxies with strong stratified correlations, but the headline numbers come from three different formulas and don't hold on natural samples; worth reviewing but needs a major revision. read the letter →

arxiv 2411.15290 v1 pith:TPZBPZ6E submitted 2024-11-22 cs.LG cs.NE

Gabriel Cortês , Nuno Lourenço , Penousal Machado This is my paper · ORCID
classification cs.LGcs.NE
keywords neuralarchitecturesearchzero-costproxiesgeneticprogrammingstructuredgrammaticalevolutiontraining-freeNASNATS-BenchKendallcorrelationstratifiedsampling
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

Neural architecture search normally requires training candidate networks to know which ones are good, making it slow and energy-hungry. Zero-cost proxies try to rank networks before training by scoring their weights and activations, but hand-designed proxies often fail when the candidate set includes poor networks as well as good ones. GreenMachine evolves proxy formulas with grammar-based genetic programming, using stratified samples of the NATS-Bench search space so that a proxy only scores well if it separates low-performing from high-performing architectures. On stratified validation sets, its best evolved proxies reach Kendall correlations of 0.89 on CIFAR-10 and 0.77 on CIFAR-100 for the size search space, and 0.78 and 0.71 for the topology search space, beating every compared zero-cost proxy. This points toward training-free architecture ranking that is cheaper and works across the full performance range.

What carries the argument

The search engine is grammar-based genetic programming using Structured Grammatical Evolution, which represents each proxy as a syntax tree over per-layer statistics collected from a randomly initialized network: weights and gradients, plus forward and backward activations after a batch of random data, a real data batch, or a noisy batch. Each tree maps those tensors through arithmetic, norm, activation, comparison, and similarity operations into a single score, and the network score is the mean over layers. Fitness is the sum of absolute Kendall $\tau$ correlations across search spaces and datasets. The other load-bearing component is stratified sampling: the search space is divided into five accuracy bins and the evaluation set is built to represent all bins, so proxies are rewarded for ordering poor networks below good ones instead of exploiting the natural skew toward high accuracy.

What would settle it

Score a new search space such as NAS-Bench-201 with the best GreenMachine proxy and compute the Kendall correlation against true test accuracy; if it drops to the level of a random score or below existing zero-cost proxies, the generalization claim fails. A second test is to re-run evolution with 120 randomly sampled architectures instead of stratified ones and check whether the stratified validation correlations collapse.

Watch

Extended reading notes

Core claim

The paper's central claim is that zero-cost proxies discovered by evolution outperform existing zero-cost proxy methods specifically when the task is to distinguish low- and high-performing networks. The authors demonstrate this by evaluating the evolved formulas on NATS-Bench using stratified sampling that covers the full accuracy range. Their best solutions achieve mean absolute Kendall $\tau$ correlations of 0.89 on CIFAR-10 and 0.77 on CIFAR-100 in the size search space (NATS-Bench-SSS), and 0.78 on CIFAR-10 and 0.71 on CIFAR-100 in the topology search space (NATS-Bench-TSS), exceeding all compared baselines on stratified validation sets. On randomly sampled validation sets, where most networks are high-performing and similar to one another, the advantage largely disappears; the authors interpret this as showing that stratified evaluation is what makes the ranking capability visible and useful.

Load-bearing premise

The method assumes that fitness computed on only 120 sampled architectures is a reliable guide to performance across the full search space, and the paper does not specify how those 120 networks are sampled; if that sample is unrepresentative, the evolved proxies may not generalize beyond the benchmark.

Editorial extensions

If this is right

  • Training-free architecture search can become substantially more reliable on broad search spaces, since the evolved proxies rank unseen networks across the full accuracy range.
  • Automatic proxy discovery is a viable alternative to hand-designed proxies: the best GreenMachine solutions beat all compared baselines on stratified NATS-Bench sets.
  • Because scoring does not require training networks, architecture search with such proxies consumes much less energy, which is the sustainability motivation of the paper.
  • Evaluation protocol matters: comparisons on randomly sampled networks can mask a proxy's ability to separate good and bad architectures, so stratified correlation should be reported alongside random-sampling results.

Reading between the lines

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

  • Beyond the reported results, the same evolutionary pipeline should be tested on other search spaces and datasets, since the paper validates only on NATS-Bench.
  • An implicit methodological point is that the 120-network fitness sample may itself need to be stratified; the paper describes stratification for evaluation but does not specify how the 120 training architectures are chosen, so the contribution of sampling versus the evolved formulas is not cleanly separated.
  • A stronger downstream test would run an actual architecture search with the evolved proxy and compare the final networks' accuracy with searches guided by existing proxies, rather than only comparing ranking correlations.
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

3 major / 5 minor

Summary. GreenMachine proposes an evolutionary (structured grammatical evolution) approach to automatically synthesize zero-cost proxies for neural architecture performance prediction. Proxies are scored by Kendall's tau between proxy output and test accuracy on samples from NATS-Bench (TSS and SSS; CIFAR-10, CIFAR-100, ImageNet16-120). The fitness function is the sum of absolute Kendall correlations over six search-space/dataset combinations. The paper reports that on stratified validation sets, the discovered proxies outperform existing zero-cost proxies, with headline Kendall correlations of 0.89 (SSS CIFAR-10), 0.77 (SSS CIFAR-100), 0.78 (TSS CIFAR-10), and 0.71 (TSS CIFAR-100). It also reports results on non-stratified validation sets, where the advantage is much weaker or reversed. The manuscript concludes that the discovered proxies perform better than existing methods when distinguishing between low- and high-performing networks.

Significance. If the central claim survives scrutiny, the paper offers a practical way to automate the design of training-free NAS proxies and would strengthen the evidence that evolved, parameter-free formulas can rival hand-designed zero-cost proxies. Positive features of the manuscript are that the code is made publicly available, the evolved formulas contain no fitted numeric constants, and the validation sets appear to be disjoint from the networks used during evolution. The significance is limited, however, by the fact that all experiments are confined to NATS-Bench and by the evaluation-protocol issues discussed below, which currently prevent the abstract's headline claim from being attributed to a single, well-specified GreenMachine solution.

major comments (3)
  1. [Abstract and Section 4, Tables 3 and 4] The headline correlations are not produced by a single GreenMachine proxy. In Table 4, the 0.89 on SSS CIFAR-10 belongs to GM-3, the 0.77 on SSS CIFAR-100 belongs to GM-2, and the 0.78 and 0.71 on TSS CIFAR-10 and CIFAR-100 belong to GM-1. The manuscript does not explain how GM-1, GM-2, and GM-3 were selected after the five evolutionary runs (best per run, best per benchmark, or best per dataset). As written, the abstract and conclusion attribute these numbers to 'our approach' or 'the zero-cost proxies discovered by our approach', which invites post-hoc selection. Please report the full 3x6 correlation matrix for each of GM-1, GM-2, and GM-3, and either specify an a-priori selection rule or explicitly reframe the claim as 'for each benchmark/dataset a different evolved proxy can be chosen'.
  2. [Section 4, Tables 3 and 4; Section 5] The claimed superiority is established only under stratified validation sampling, which is not the distribution a NAS practitioner encounters. On the non-stratified sets (Table 3), EZNAS is better than all GreenMachine proxies on TSS for all three datasets (e.g., 61.1 vs 50.2 on CIFAR-10), and Synflow beats GM-3 on SSS CIFAR-10 (76.8 vs 75.6) and beats GM-2 on SSS ImageNet16-120 (79.5 vs 78.0). The abstract qualifies the results as being on the stratified strategy, but the conclusion generalizes to 'distinguishing between low- and high-performing networks' without noting that this capability is demonstrated only when the validation set is deliberately balanced across accuracy bins. Since test-accuracy stratification uses ground-truth labels that are unknown at proxy-deployment time, the paper should either restrict the central claim to the stratified diagnostic protocol or provide transfer evidence, such as top-k rank agreement, search-space-level ranking performance, or downstream NAS search quality on naturally sampled networks.
  3. [Section 3.2.1 and Section 3.2.5] The fitness-evaluation sampling is not fully specified, and the description of the validation sets is ambiguous. Section 3.2.1 describes stratifying the search space into five accuracy bins and sampling 20 networks per dataset per search space, which yields 120 networks (consistent with Table 1), but it does not state how those 120 networks are split among the five bins, whether the same 120 networks are used for every generation/run, or what random seeds are used. Section 4 says validation is performed on '4,500 new networks' and also says '30 sets of 150 networks from each search space and dataset'; these two statements are arithmetically inconsistent unless '4,500' refers to total unique networks and the per-condition sets are overlapping. The paper should state explicitly how the 120 fitness networks are sampled, confirm that the validation sets exclude all 120 fitness networks, and report the exact number of unique validation networks per search-space/dataset condition.
minor comments (5)
  1. [Section 4, last paragraph before Section 5] The text says 'see Tab. 3' when discussing the results obtained with stratified sampling, but those results are in Table 4; please correct the cross-reference.
  2. [Section 3.2.3 and Figure 7 / Table 6] The terminal symbols used in the evolved formulas, such as pass_noise_wt, pass_perturbation_fwd_output, and random_grad, are not defined in the main text or in the supplementary material. Please provide a complete list of terminals and their definitions alongside the feature-extraction description.
  3. [Section 3.2.3] The feature-extraction description says 20 features are extracted per layer but does not enumerate them. Please list the 20 features explicitly, or state that the full list is in the supplementary material if it is.
  4. [Section 4] The phrase '4,500 new networks' is ambiguous, as noted above; please clarify whether this is the total number of unique validation networks or the number of sets times set size, and give per-condition counts.
  5. [Tables 3 and 4] The paper reports means and standard deviations over 30 sets but no significance tests; for key comparisons, such as GM-3 (88.8 ± 0.9) versus EZNAS (82.8 ± 0.9) on SSS CIFAR-10 in Table 4, please report a test or a confidence interval to support the claim of superiority.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline correlations are computed on held-out NATS-Bench networks, not on the fitness sample or by construction.

full rationale

The paper's derivation chain is an evolutionary search: proxy formulas are randomly generated and selected using a fitness function (absolute Kendall tau over 120 stratified NATS-Bench networks, Section 3.2.5), and the resulting formulas are then evaluated on disjoint validation sets of 150 networks under both stratified and non-stratified sampling (Section 4). The reported correlations are therefore out-of-sample with respect to the fitness data, and the proxy scores are not defined in terms of the accuracies they predict. The only self-citations are to the authors' prior work on Structured Grammatical Evolution [24, 25], used as the representation and mutation/recombination engine; this is a methodological tool with independent published support and is not used to forbid alternatives or to import a uniqueness result. The skeptical concern that stratified evaluation oversamples low-performing networks and that GreenMachine does not consistently beat baselines on natural random samples is a generalizability/external-validity issue, not a circularity issue: the abstract and tables honestly restrict the superiority claim to the stratified protocol. No fitted numeric constants are introduced, no prediction is a renamed input, and no load-bearing premise is justified solely by self-citation. Hence no circular step can be exhibited from the paper's own equations.

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

No numeric constants are fitted within the discovered proxy formulas; the evolutionary search optimizes discrete tree structures, so the free-parameter list is empty. The method relies on several domain assumptions and paper-specific design choices, captured above. The evolved proxy formulas themselves, GM-1, GM-2, and GM-3, are outputs of the search, not independent postulated entities.

assumptions (5)
  • domain assumption NATS-Bench test accuracies are treated as exact ground truth for architecture quality.
    Used to compute every Kendall correlation in Sections 3.2.2, 4, and the validation. The benchmark's accuracy values are not questioned, and all proxy scores are judged against them.
  • domain assumption Features extracted from randomly initialized networks with one batch of data (random, clean, or perturbed) are informative about final trained accuracy.
    Adopted from the zero-cost proxy literature, including NASWOT, Synflow, and ZiCo. Section 3.2.3 builds the terminal set on this premise without independent justification in this paper.
  • ad hoc to paper The fitness function, the sum of absolute Kendall correlations across all search space and dataset combinations, is a valid objective for discovering general proxies.
    Defined in Section 3.2.2. This objective couples all tasks and uses absolute values, a design choice not derived from theory and not compared against alternative objectives.
  • ad hoc to paper Stratifying architectures into five bins by test accuracy yields a representative training and evaluation distribution.
    Introduced in Section 3.2.1 to counter skew in the search space. The number of bins, the binning boundaries, and the sample size of 150 networks are chosen without sensitivity analysis.
  • domain assumption The feature vocabulary maps to reproducible terminal variables such as pass_noise_wt and pass_perturbation_bwd_input.
    Section 3.2.3 and the formulas in Figure 7 rely on these terminals, but the exact computation of each terminal from the archived per-layer statistics is not fully specified in the manuscript.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GreenMachine: Automatic Design of Zero-Cost Proxies for Energy-Efficient NAS." pith.science (2026). https://pith.science/paper/TPZBPZ6E

@misc{pith2026241115290,
  author       = {Pith},
  title        = {Pith review of: GreenMachine: Automatic Design of Zero-Cost Proxies for Energy-Efficient NAS},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TPZBPZ6E}},
  note         = {Machine review of arXiv:2411.15290}
}
read the original abstract

Artificial Intelligence (AI) has driven innovations and created new opportunities across various sectors. However, leveraging domain-specific knowledge often requires automated tools to design and configure models effectively. In the case of Deep Neural Networks (DNNs), researchers and practitioners usually resort to Neural Architecture Search (NAS) approaches, which are resource- and time-intensive, requiring the training and evaluation of numerous candidate architectures. This raises sustainability concerns, particularly due to the high energy demands involved, creating a paradox: the pursuit of the most effective model can undermine sustainability goals. To mitigate this issue, zero-cost proxies have emerged as a promising alternative. These proxies estimate a model's performance without the need for full training, offering a more efficient approach. This paper addresses the challenges of model evaluation by automatically designing zero-cost proxies to assess DNNs efficiently. Our method begins with a randomly generated set of zero-cost proxies, which are evolved and tested using the NATS-Bench benchmark. We assess the proxies' effectiveness using both randomly sampled and stratified subsets of the search space, ensuring they can differentiate between low- and high-performing networks and enhance generalizability. Results show our method outperforms existing approaches on the stratified sampling strategy, achieving strong correlations with ground truth performance, including a Kendall correlation of 0.89 on CIFAR-10 and 0.77 on CIFAR-100 with NATS-Bench-SSS and a Kendall correlation of 0.78 on CIFAR-10 and 0.71 on CIFAR-100 with NATS-Bench-TSS.

Figures

Figures reproduced from arXiv: 2411.15290 by the authors.

Figure 3
Figure 3. Example of Grammatical Evolution mapping. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 1
Figure 1. Simplified visualization of the flow of an EA. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Example of a grammar. Despite GE’s flexibility and ease of use, it has suf￾fered from issues, such as low locality and high redun￾dancy [24]. To address these limitations, Structured Grammatical Evolution (SGE) [25] was introduced. SGE mitigates these issues by associating each gene Genotype: [46, 15, 17, 28, 50, 42, 22, 19, 51, 35] <expr> <trig>(<expr>) 46 mod 3 = 1 cos(<expr>) 15 mod 2 = 1 cos(<var>) 17 mod 3 = 2 … view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Histograms of test accuracy for the NATS-Bench [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Evolution of the Kendall correlation coefficient on [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Representation of the GreenMachine-2 solution, [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Formulas of the best solutions found. 12 [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 37 canonical work pages

  1. [1]

    Abdelfattah, Abhinav Mehrotra, Lukasz Dudziak, and Nicholas Donald Lane

    Mohamed S. Abdelfattah, Abhinav Mehrotra, Lukasz Dudziak, and Nicholas Donald Lane. Zero-cost proxies for lightweight NAS. In 9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Aus- tria, May 3-7, 2021. OpenReview.net, 2021. 4, 8, 13

    work page 2021

  2. [2]

    EZNAS: evolving zero-cost proxies for neural ar- chitecture scoring

    Yash Akhauri, Juan Pablo Mu˜noz, Nilesh Jain, and Ravi Iyer. EZNAS: evolving zero-cost proxies for neural ar- chitecture scoring. In Advances in Neural Information Processing Systems 35: Annual Conference on Neural In- formation Processing Systems 2022, NeurIPS 2022, New Orleans, LA, USA, November 28 - December 9, 2022, 2022. 4, 8, 13

    work page 2022

  3. [3]

    Plant disease detection using ai based vgg-16 model.Interna- tional Journal of Advanced Computer Science and Applica- tions, 2022

    Anwar Abdullah Alatawi, Shahd Maadi Alomani, Najd Ibrahim Alhawiti, and Muhammad Ayaz. Plant disease detection using ai based vgg-16 model.Interna- tional Journal of Advanced Computer Science and Applica- tions, 2022. 1

    work page 2022

  4. [4]

    S. A. Alowais, S. S. Alghamdi, N. Alsuhebany, T. Alqahtani, A. I. Alshaya, S. N. Almohareb, A. Al- dairem, M. Alrashed, K. Bin Saleh, H. A. Badreldin, M. S. Al Yami, S. Al Harbi, and A. M. Albekairy. Revo- lutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Medical Education , 23(1):689,

    work page

  5. [5]

    Low-Cost Recurrent Neural Network Expected Performance Evaluation

    Andr ´es Camero, Jamal Toutouh, and Enrique Alba. Low-cost recurrent neural network expected perfor- mance evaluation. CoRR, abs/1805.07159, 2018. 3

    work page Pith review arXiv 2018

  6. [6]

    Neural architecture search on imagenet in four GPU hours: A theoretically inspired perspective

    Wuyang Chen, Xinyu Gong, and Zhangyang Wang. Neural architecture search on imagenet in four GPU hours: A theoretically inspired perspective. In 9th In- ternational Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021 . OpenRe- view.net, 2021. 4, 8, 13

    work page 2021

  7. [7]

    Towards physical plausibility in neu- roevolution systems

    Gabriel Cort ˆes, Nuno Lourenc ¸o, and Penousal Machado. Towards physical plausibility in neu- roevolution systems. In Applications of Evolutionary Computation - 27th European Conference, EvoApplica- tions 2024, Held as Part of EvoStar 2024, Aberystwyth, UK, April 3-5, 2024, Proceedings, Part II , pages 76–90. Springer, 2024. 2

    work page 2024

  8. [8]

    Threats and opportunities with ai-based cyber security intrusion detection: A review

    Bibhu Dash, Md Meraj Ansari, Pawankumar Sharma, and Azad Ali. Threats and opportunities with ai-based cyber security intrusion detection: A review. Inter- national Journal of Software Engineering & Applications ,

    work page

Show all 41 references
  1. [9]

    The growing energy footprint of artificial intelligence

    Alex de Vries. The growing energy footprint of artificial intelligence. Joule, 7(10):2191–2194, 2023. 1

    2023

  2. [10]

    Parzc: Parametric zero-cost proxies for efficient NAS

    Peijie Dong, Lujun Li, Xinglin Pan, Zimian Wei, Xi- ang Liu, Qiang Wang, and Xiaowen Chu. Parzc: Parametric zero-cost proxies for efficient NAS. CoRR, abs/2402.02105, 2024. 2

    2024 arXiv

  3. [11]

    NAS-Bench-201: Extending the scope of reproducible neural architecture search

    Xuanyi Dong and Yi Yang. NAS-Bench-201: Extending the scope of reproducible neural architecture search. In 8th International Conference on Learning Representa- tions, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020. OpenReview.net, 2020. 4

    2020

  4. [12]

    NATS-Bench: Benchmarking NAS algorithms for architecture topology and size

    Xuanyi Dong, Lu Liu, Katarzyna Musial, and Bogdan Gabrys. NATS-Bench: Benchmarking NAS algorithms for architecture topology and size. IEEE Trans. Pattern Anal. Mach. Intell., 44(7):3634–3646, 2022. 4, 5

    2022

  5. [13]

    A. E. Eiben and James E. Smith. Introduction to Evolu- tionary Computing, Second Edition. Springer, 2015. 2

    2015

  6. [14]

    DEAP: evolutionary algorithms made easy

    F ´elix-Antoine Fortin, Franc ¸ois-Michel De Rainville, Marc-Andr´e Gardner, Marc Parizeau, and Christian Gagn´e. DEAP: evolutionary algorithms made easy. J. Mach. Learn. Res., 13:2171–2175, 2012. 4

    2012

  7. [15]

    Ai and the econ- omy

    Jason Furman and Robert Seamans. Ai and the econ- omy. Working Paper 24689, National Bureau of Eco- nomic Research, 2018. 1

    2018

  8. [16]

    I, robot: the three laws of robotics and the ethics of the peopleless economy

    Orlando Gomes. I, robot: the three laws of robotics and the ethics of the peopleless economy. AI Ethics, 4 (2):257–272, 2024. 1

    2024

  9. [17]

    2024 Environmental Report

    Google. 2024 Environmental Report. https : //sustainability.google/reports/google- 2024- environmental- report/, 2024. [Accessed 05-10-2024]. 1

    2024

  10. [18]

    Deivalaskhmi, G

    Muhammad Hamis Haider, Hao Zhang, S. Deivalaskhmi, G. Lakshmi Narayanan, and Seok-Bum Ko. Is Neuromorphic Computing the Key to Power-Efficient Neural Networks: A Survey , pages 91–113. Springer Nature Switzerland, Cham, 2024. 1

    2024

  11. [19]

    Diamos, Heewoo Jun, Hassan Kianinejad, Md

    Joel Hestness, Sharan Narang, Newsha Ardalani, Gre- gory F. Diamos, Heewoo Jun, Hassan Kianinejad, Md. Mostofa Ali Patwary, Yang Yang, and Yanqi Zhou. Deep learning scaling is predictable, empirically.CoRR, abs/1712.00409, 2017. 6

    2017 arXiv

  12. [20]

    Bias in AI-Based Decision-Making , pages 275–

    Adheesh Kadiresan, Yuvraj Baweja, and Obi Og- banufe. Bias in AI-Based Decision-Making , pages 275–

  13. [21]

    AZ-NAS: assem- bling zero-cost proxies for network architecture search

    Junghyup Lee and Bumsub Ham. AZ-NAS: assem- bling zero-cost proxies for network architecture search. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 5893–5903. IEEE, 2024. 2

    2024

  14. [22]

    Zico: Zero-shot NAS via inverse co- efficient of variation on gradients

    Guihong Li, Yuedong Yang, Kartikeya Bhardwaj, and Radu Marculescu. Zico: Zero-shot NAS via inverse co- efficient of variation on gradients. In The Eleventh In- ternational Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023 . OpenReview.net,

    2023

  15. [23]

    Zen-nas: A zero-shot NAS for high-performance image recogni- tion

    Ming Lin, Pichao Wang, Zhenhong Sun, Hesen Chen, Xiuyu Sun, Qi Qian, Hao Li, and Rong Jin. Zen-nas: A zero-shot NAS for high-performance image recogni- tion. In 2021 IEEE/CVF International Conference on Com- puter Vision, ICCV 2021, Montreal, QC, Canada, October 10-17, 2021, p...

    2021

  16. [24]

    Pereira, and Ernesto Costa

    Nuno Lourenc ¸o, Francisco B. Pereira, and Ernesto Costa. SGE: A structured representation for grammat- ical evolution. In Artificial Evolution - 12th International Conference, Evolution Artificielle, EA 2015, Lyon, France, October 26-28, 2015. Revised Selected Papers , pages 136–

    2015

  17. [25]

    Pereira, Ernesto Costa, and Penousal Machado

    Nuno Lourenc ¸o, Filipe Assunc ¸˜ao, Francisco B. Pereira, Ernesto Costa, and Penousal Machado. Structured grammatical evolution: A dynamic approach. In Hand- book of Grammatical Evolution, pages 137–161. Springer,

  18. [26]

    Networks of spiking neurons: The third generation of neural network models.Neural Net- works, 10(9):1659–1671, 1997

    Wolfgang Maass. Networks of spiking neurons: The third generation of neural network models.Neural Net- works, 10(9):1659–1671, 1997. 1

    1997

  19. [27]

    Neural architecture search without training

    Joe Mellor, Jack Turner, Amos Storkey, and Elliot J Crowley. Neural architecture search without training. In Proceedings of the 38th International Conference on Ma- chine Learning, pages 7588–7598. PMLR, 2021. 4, 8, 13

    2021

  20. [28]

    Patterson, Joseph Gonzalez, Urs H ¨olzle, Quoc V

    David A. Patterson, Joseph Gonzalez, Urs H ¨olzle, Quoc V . Le, Chen Liang, Lluis-Miquel Munguia, Daniel Rothchild, David R. So, Maud Texier, and Jeff Dean. The carbon footprint of machine learning training will plateau, then shrink. Computer, 55(7):18–28, 2022. 1

    2022

  21. [29]

    Ramirez, Song-Kyoo Kim, Hussam M

    Miguel A. Ramirez, Song-Kyoo Kim, Hussam M. N. Al Hamadi, Ernesto Damiani, Young-Ji Byon, Tae-Yeon Kim, Chung-Suk Cho, and Chan Yeob Yeun. Poisoning attacks and defenses on artificial intelligence: A survey. CoRR, abs/2202.10276, 2022. 1

    2022 arXiv

  22. [30]

    Microsoft’s carbon emissions up nearly 30 https://www.msn.com/en-us/money/other/ microsofts - carbon - emissions - up - nearly - 30-thanks-to-ai/ar-BB1mvgao , 2024

    Dan Robinson. Microsoft’s carbon emissions up nearly 30 https://www.msn.com/en-us/money/other/ microsofts - carbon - emissions - up - nearly - 30-thanks-to-ai/ar-BB1mvgao , 2024. [Accessed 05-10-2024]. 1

    2024

  23. [31]

    A comprehensive survey on model quantiza- tion for deep neural networks in image classification

    Babak Rokh, Ali Azarpeyvand, and Alireza Khantey- moori. A comprehensive survey on model quantiza- tion for deep neural networks in image classification. ACM Trans. Intell. Syst. Technol., 14(6), 2023. 1

    2023

  24. [32]

    Conor Ryan, JJ Collins, and Michael O. Neill. Gram- matical evolution: Evolving programs for an arbitrary language. In Genetic Programming, pages 83–96, Berlin, Heidelberg, 1998. Springer Berlin Heidelberg. 2

    1998

  25. [33]

    Assessment of problem modality by differ- ential performance of lexicase selection in genetic pro- gramming: a preliminary report

    Lee Spector. Assessment of problem modality by differ- ential performance of lexicase selection in genetic pro- gramming: a preliminary report. In Genetic and Evo- lutionary Computation Conference, GECCO ’12, Philadel- phia, P A, USA, July 7-11, 2012, Companion Material Pro- c...

    2012

  26. [34]

    Unemployment in the AI age

    Grace Su. Unemployment in the AI age. AI Matters, 3 (4):35–43, 2018. 1

    2018

  27. [35]

    Hidenori Tanaka, Daniel Kunin, Daniel L. K. Yamins, and Surya Ganguli. Pruning neural networks without any data by iteratively conserving synaptic flow. InAd- vances in Neural Information Processing Systems 33: An- nual Conference on Neural Information Processing Systems 2020,...

    2020

  28. [36]

    AI and ML for Transportation Route Op- timization

    Rama Vaddy. AI and ML for Transportation Route Op- timization. International Transactions in Machine Learn- ing, 5(5):1–19, 2023. 1

    2023

  29. [37]

    Sustainable AI: AI for sus- tainability and the sustainability of AI

    Aimee van Wynsberghe. Sustainable AI: AI for sus- tainability and the sustainability of AI. AI Ethics, 1(3): 213–218, 2021. 1

    2021

  30. [38]

    Willis, H.G

    M.-J. Willis, H.G. Hiden, P . Marenbach, B. McKay, and G.A. Montague. Genetic programming: an introduc- tion and survey of applications. In Second International Conference On Genetic Algorithms In Engineering Systems: Innovations And Applications, pages 314–319, 1997. 2

    1997

  31. [39]

    Training-free neu- ral architecture search: A review

    Meng-Ting Wu and Chun-Wei Tsai. Training-free neu- ral architecture search: A review. ICT Express, 10(1): 213–231, 2024. 3

    2024

  32. [40]

    A survey of resource- efficient LLM and multimodal foundation models

    Mengwei Xu, Wangsong Yin, Dongqi Cai, Rongjie Yi, Daliang Xu, Qipeng Wang, Bingyang Wu, Yihao Zhao, Chen Yang, Shihe Wang, Qiyang Zhang, Zhenyan Lu, Li Zhang, Shangguang Wang, Yuanchun Li, Yunxin Liu, Xin Jin, and Xuanzhe Liu. A survey of resource- efficient LLM and multimodal...

    2024 arXiv

  33. [285]

    Springer International Publishing, Cham, 2022. 1

    2022

Pith tools

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