Pith. sign in

REVIEW 5 major objections 6 minor 38 references

An Introduction to Advanced Machine Learning : Meta Learning Algorithms, Applications and Promises

T0 review · 5 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read This chapter argues that meta-learning, or learning to learn, is the advanced machine learning paradigm that can solve few-shot, one-shot, and zero-shot learning by reusing prior experience.

desk verdict A broad but careless meta-learning survey: useful only as an entry-level draft, and unreliable due to misattributions and a garbled equation. read the letter →

arxiv 1908.09788 v1 pith:7M7HS4CK submitted 2019-08-26 cs.LG stat.ML

Farid Ghareh Mohammadi , M. Hadi Amini , Hamid R. Arabnia This is my paper · ORCID
classification cs.LGstat.ML
keywords meta-learninglearningtolearnfew-shotone-shotzero-shotmodel-basedmetric-basedgradientdescent
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that classical machine learning, trained on static offline batches, cannot cope with online, distributed data that offers only a few labeled samples per class. It presents meta-learning (MTL), the process of learning to learn, as the promising alternative and organizes the field into three algorithm families: model-based, metric-based, and gradient-descent-based. On this map, MTL delivers three capabilities: few-shot learning (k samples per class), one-shot learning (one sample per class), and zero-shot learning (no samples, using semantic side information). The reader is meant to come away seeing MTL as the route to autonomous agents that adapt quickly to new tasks from past experience. Because this is an introductory survey, the claim is organizational and promissory rather than a new experimental result.

What carries the argument

The organizing device is a three-way taxonomy of meta-learning algorithms. Model-based MTL trains a recurrent or memory-augmented network that ingests data sequentially and updates hyper-parameters from a few examples; metric-based MTL learns a comparison space and classifies by distance between encoded inputs; gradient-descent-based MTL adjusts parameters so that one or a few gradient steps generalize to a new task, MAML being the canonical case. The taxonomy does the argumentative work: Table 2 places more than twenty cited algorithms into cells of this three-by-three promise structure, so the whole field appears to converge on few-shot, one-shot, and zero-shot learning. The meta-training / meta-testing layer distinction then frames how these promises are realized.

What would settle it

Run the datasets the cited few-shot papers use (for instance Omniglot and miniImageNet at 1-shot and 5-shot) across all three taxonomy families and compare against plain fine-tuning of a pretrained network; if none of the three families beats that baseline on held-out classes, the claim that MTL unlocks low-shot learning is contradicted.

Watch

Extended reading notes

Core claim

The central claim is that meta-learning (MTL) is the advanced machine learning paradigm for emerging problems: an agent that learns to learn can use past tasks to choose a non-fixed inductive bias and adapt quickly to a new task, unlike base learning where the bias is fixed in advance. The paper partitions MTL research into model-based algorithms (networks that update their own hyper-parameters, typically recurrent or memory-augmented), metric-based algorithms (similarity learners such as Siamese and matching networks), and gradient-descent-based algorithms (optimization-based meta-learners, with MAML, Model-Agnostic Meta-Learning, as the flagship). It then attaches these families to three promises: few-shot learning, one-shot learning as its special case, and zero-shot learning via mapping from a semantic space to a vector space. It also motivates an online MTL setting in which past experience is exploited while tasks arrive sequentially, citing an existing regret-based meta-learner. The chapter's contribution is this three-way taxonomy and its promise map, not a new algorithm or benchmark.

Load-bearing premise

The argument rests on the assumption that every important meta-learning method fits into one of three categories—model-based, metric-based, or gradient-descent-based—and that the papers selected to represent those categories actually deliver the promised few-shot, one-shot, and zero-shot abilities.

Editorial extensions

If this is right

  • If the promise holds, deployed systems can recognize new classes from a handful of labeled examples, which matters wherever labeling is expensive or data arrives fast.
  • One-shot learning becomes the k=1 special case of few-shot learning, so progress in few-shot benchmarks should transfer directly to one-shot settings.
  • Zero-shot learning would let an agent classify classes it has never seen by exploiting semantic descriptions of those classes.
  • An online MTL agent would combine sequential task arrival with experience reuse, giving a path from the current one-task-at-a-time learning paradigm toward continual lifelong learning.

Reading between the lines

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

  • Because this is an introductory survey, the experimental weight behind its promises lives in the cited papers; a reader who wants to verify the promises should examine those evaluations directly.
  • The chapter motivates online MTL but does not give it a formal objective; defining it as regret minimization over sequentially revealed tasks with meta-learned priors would make the promise testable.
  • The taxonomy leaves no explicit cell for hybrids that combine metric similarity with gradient-based adaptation; checking whether such hybrids beat all three pure families on common few-shot benchmarks would test the taxonomy's completeness.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. This manuscript, an arXiv preprint labeled as a chapter, presents an introductory survey of meta-learning (MTL). It organizes MTL into model-based, metric-based, and gradient-descent-based categories and discusses few-shot learning, one-shot learning, and zero-shot learning as the main promised applications. The abstract states that the chapter introduces emerging challenges and investigates how autonomous agents can learn to learn. The paper relies on prior work by the same authors [1,2] for background on optimization and on external foundational works, such as MAML, Matching Networks, and the LSTM meta-learner, for the central content. It also states that it motivates and presents the online MTL problem setting, but the submitted text does not contain a new formalization or substantive discussion of that setting.

Significance. The paper is positioned as a pedagogical survey rather than as an original research contribution. If its taxonomy and table of methods were accurate, it could serve as a useful introduction; it gathers a set of influential meta-learning references and groups them into three families, and it does not reuse fitted parameters or derivations from cited work in a circular way. However, the survey's significance is currently compromised by several checkable errors: a garbled task-distribution equation, misattributed methods in Table 2, and at least one statement in Section 3.3 that contradicts the behavior of the methods being cited. These errors directly affect the central claim that the proposed taxonomy covers the field and that MTL promises to solve emerging few-shot, one-shot, and zero-shot learning problems. The paper offers no original evidence for these promise claims, so as written it cannot be considered a reliable survey.

major comments (5)
  1. [Section 4, equation for Ti] The displayed equation "Ti = Ík i=1(L(xi ,/y.alti), q(xi), q(xt +1|xt , qt))" is not a valid or meaningful expression; the symbols "/y.alti" and the mixing of a loss function L with an input distribution q(x_i) and a transition q(x_{t+1}|x_t,q_t) do not correspond to a standard task distribution, and the expression is not a faithful rendering of the corresponding notation in Finn et al. [20]. Because this equation is the only formal specification of the few-shot/one-shot task setting, it should be replaced with a correct definition, such as a task T_i consisting of a sampled set of input-output pairs with a task distribution, or the passage should explicitly quote the relevant equation from [20].
  2. [Table 2, row [18]] The row for Koch et al. [18] lists the proposed method as "MIL" (Meta Imitation Learning), but [18] is the Siamese-network paper for one-shot image recognition; MIL is the method of Finn et al. [28]. This is a concrete misattribution in the central survey table and needs to be corrected.
  3. [Table 2, row [27]] The row for "Low-shot learning from imaginary data" [27] is classified as metric-based. That paper generates synthetic training examples from a hallucination network and does not learn a metric space, so grouping it with metric-based methods misrepresents its mechanism. The paper should be reclassified, or an explicit justification for the metric-based label should be provided.
  4. [Section 3.3] The sentence stating that gradient-descent-based MTL "does not rely on small number of samples" is the opposite of what the cited methods do: Ravi and Larochelle [4] and Finn et al. [20] are explicitly few-shot learning methods designed for fast adaptation from small support sets. This claim should be corrected or removed.
  5. [Sections 1, 4.1, and 4.3] The introduction promises that the chapter will "motivate and present the online MTL problem setting," but no such setting is formalized anywhere in the text; the only mention of online MTL is a citation to Finn et al. [36] in Section 4.1. Similarly, Section 4.3 asserts without evidence that ZSL "promised to yield a result better than FSL and OSL," even though ZSL is a different problem setting rather than a method that dominates the other two. Either add the promised formulation and supporting evidence, or delete these claims.
minor comments (6)
  1. [Throughout] "Gradient decent" should read "gradient descent" in Section 3.3, Table 2, and figure labels.
  2. [Section 4] There are several typos: "equitation" should be "equation," "bootstraping" should be "bootstrapping," "converg" should be "converge," and "decent work" should be "considerable work."
  3. [Table 2] The table lists the zero-shot visual imitation paper by Pathak et al. twice, as [33] and [34], with different venues; the authors should verify whether these are distinct publications or duplicates.
  4. [Section 2] The notation "Pi x D -> M" is unexplained; the symbols should be defined or the formula removed.
  5. [Figures 7-9] Figures 7-9 are referenced in the text but their content is not described or interpreted sufficiently for a stand-alone survey; explanatory captions and discussion should be added.
  6. [Table 1] The abbreviation list uses "fSL" for few-shot learning while the text uses "FSL"; the notation should be made consistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is a literature survey whose self-citations are background material and whose claims rest on external cited works.

full rationale

This is an introductory survey, not a derivation paper. The abstract references the authors' own earlier work [1,2] on optimization for offline data, but that material is used only to motivate the survey's subject, meta-learning, and is never invoked to justify a prediction or a formal result. The paper's central statements, that meta-learning enables few-shot, one-shot, and zero-shot learning, are supported by citations to external works such as MAML [20], matching networks [19], and online meta-learning [36]; those citations are independent evidence rather than a self-citation chain. There are no fitted parameters, no equations that are defined in terms of the quantity they purport to predict, and no uniqueness theorem imported from the authors' prior work. The garbled equation in Section 4 and the Table 2 misclassifications, such as Koch et al. labelled MIL, are accuracy and quality problems, not circularity: they do not make the survey's claims true by construction. The paper's phrasing in Section 1 that it will 'present the online MTL problem setting' is not backed by a new formulation, but the paper itself cites Finn et al. [36] for online meta-learning, so this is a novelty or exposition issue rather than a circularity. Thus no circular step meets the evidentiary standard required by the review rules, and the appropriate finding is no significant circularity.

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

The paper's claims rest on the authors' subjective selection and interpretation of prior work. No new entities, free parameters, or experimental data are introduced.

assumptions (1)
  • domain assumption The selected literature in Table 2 is representative of the entire meta-learning field and supports the proposed three-way taxonomy.
    The paper organizes meta-learning into model-based, metric-based, and gradient-descent-based categories based on a curated list of papers; no evidence for completeness is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Introduction to Advanced Machine Learning : Meta Learning Algorithms, Applications and Promises." pith.science (2026). https://pith.science/paper/7M7HS4CK

@misc{pith2026190809788,
  author       = {Pith},
  title        = {Pith review of: An Introduction to Advanced Machine Learning : Meta Learning Algorithms, Applications and Promises},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7M7HS4CK}},
  note         = {Machine review of arXiv:1908.09788}
}
read the original abstract

In [1, 2], we have explored the theoretical aspects of feature extraction optimization processes for solving largescale problems and overcoming machine learning limitations. Majority of optimization algorithms that have been introduced in [1, 2] guarantee the optimal performance of supervised learning, given offline and discrete data, to deal with curse of dimensionality (CoD) problem. These algorithms, however, are not tailored for solving emerging learning problems. One of the important issues caused by online data is lack of sufficient samples per class. Further, traditional machine learning algorithms cannot achieve accurate training based on limited distributed data, as data has proliferated and dispersed significantly. Machine learning employs a strict model or embedded engine to train and predict which still fails to learn unseen classes and sufficiently use online data. In this chapter, we introduce these challenges elaborately. We further investigate Meta-Learning (MTL) algorithm, and their application and promises to solve the emerging problems by answering how autonomous agents can learn to learn?.

Figures

Figures reproduced from arXiv: 1908.09788 by the authors.

Figure 1
Figure 1. Overall structure of this study Last decade, researchers have studied advanced research paradigms to solve learning process. They aim to to learn using prior tasks or experiences and leverage them for future learning. One of the promising paradigm is Meta-learning (MTL). Prior studies investigated MTL methods that learn to update a function or learning rule [3, 4]. MTL differs from classic machine learning with resp… view at source ↗
Figure 2
Figure 2. The general overview of Learning against emerging data [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Model-based MTL 3.2 Metric-based learning Metric-based learning leverages metric space learning, which leads to efficient data processing and is suitable for few-shot learning. Lets consider that our goal is image classification. As model-based learning tries to learn each image pixel by pixel which takes long time and time consuming, metric-based learning overcomes this limitation by leveraging comparing given two … view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Metric based MTL 3.3 Gradient decent-based learning This model of MTL is also known as optimization-based model for tuning the parameter (θ). The idea here is to leverage stochastic gradient-decent (SGD) and for new given sample, it updates the parametes to be a univer…
Figure 5
Figure 5. Figure 5: The relation among Machine Learning, Meta Learning, and Information system [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Structure of meta-learning models In second layer, FSL tries to learn k-shot tasks, which means MTL is training by leveraging k different training data set. K-shots had generated in advance before learning process have started. Thus, MTL is known as a certain type of b…
Figure 7
Figure 7. Figure 7: Few-shot learning structure 4.2 One shot learning One-shot learning (OSL) is a critical challenge in the applications of deep neural networks. OSL is special type of few-shot learning or k-shot learning in which it choose k=1 shot for training section. In other words, …
Figure 8
Figure 8. Figure 8: Machine Learning : ML, Meta-Learning: MTL, Online Machine Learning: OML, Transfer Learning: [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: A brief studies over promises of Meta-learning [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 30 canonical work pages

  1. [36]

    Online meta-learning

    Chelsea Finn, Aravind Rajeswaran, Sham Kakade, and Sergey Levine. Online meta-learning. arXiv preprint arXiv:1902.08438, 2019

    arXiv 1902

  2. [20]

    Model-agnostic meta-learning for fast adaptation of deep networks

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. Model-agnostic meta-learning for fast adaptation of deep networks. In Proceedings of the 34th International Conference on Machine Learning-Volume 70 , pages 1126–1135. JMLR. org, 2017

    work page 2017

  3. [18]

    Siamese neural networks for one-shot image recognition

    Gregory Koch, Richard Zemel, and Ruslan Salakhutdinov. Siamese neural networks for one-shot image recognition. In ICML deep learning workshop , volume 2, 2015

    work page 2015

  4. [28]

    One-shot visual imitation learning via meta-learning

    Chelsea Finn, Tianhe Yu, Tianhao Zhang, Pieter Abbeel, and Sergey Levine. One-shot visual imitation learning via meta-learning. In 1st Conference on Robot Learning (CoRL) , 2017

    work page 2017

  5. [27]

    Low-shot learning from imaginary data

    Yu-Xiong Wang, Ross Girshick, Martial Hebert, and Bharath Hariharan. Low-shot learning from imaginary data. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , June 2018

    work page 2018

  6. [4]

    Optimization as a model for few-shot learning

    Sachin Ravi and Hugo Larochelle. Optimization as a model for few-shot learning. In International Con- ference on Learning Representations (ICLR), pages 281–288, 2017

    work page 2017

  7. [1]

    F Ghareh Mohammadi, M Hadi Amini, and Hamid R. Arabnia. Evolutionary computation, optimization and learning algorithms for data science. arXiv preprint arXiv: 1908.08006 , 2019

    work page Pith review arXiv 1908

  8. [2]

    F Ghareh Mohammadi, M Hadi Amini, and Hamid R. Arabnia. Applications of nature-inspired algorithms for dimension Reduction: Enabling efficient data analytics. arXiv preprint arXiv: 1908.1908.08563 , 2019

    work page arXiv 1908

Show all 38 references
  1. [3]

    Evolutionary principles in self-referential learning

    Jurgen Schmidhuber. Evolutionary principles in self-referential learning. Diploma thesis, 1987

    1987

  2. [5]

    A perspective view and survey of meta-learning

    Ricardo Vilalta and Youssef Drissi. A perspective view and survey of meta-learning. Artificial intelligence review, 18(2):77–95, 2002

    2002

  3. [6]

    Approximation to bayes risk in repeated play

    James Hannan. Approximation to bayes risk in repeated play. Contributions to the Theory of Games , 3:97–139, 1957

    1957

  4. [7]

    Prediction, learning, and games

    Nicolo Cesa-Bianchi and Gabor Lugosi. Prediction, learning, and games . Cambridge university press, 2006

    2006

  5. [8]

    Rule-based machine learning methods for functional prediction

    Sholom M Weiss and Nitin Indurkhya. Rule-based machine learning methods for functional prediction. Journal of Artificial Intelligence Research , 3:383–403, 1995

    1995

  6. [9]

    Advances in electronic phenotyping: from rule-based definitions to machine learning models

    Juan M Banda, Martin Seneviratne, Tina Hernandez-Boussard, and Nigam H Shah. Advances in electronic phenotyping: from rule-based definitions to machine learning models. Annual review of biomedical data science, 1:53–68, 2018

    2018

  7. [10]

    Artificial neural networks-based machine learning for wireless networks: A tutorial

    Mingzhe Chen, Ursula Challita, Walid Saad, Changchuan Yin, and M´ erouane Debbah. Artificial neural networks-based machine learning for wireless networks: A tutorial. IEEE Communications Surveys & Tutorials, 2019

    2019

  8. [11]

    Cost-sensitive support vector machines

    Arya Iranmehr, Hamed Masnadi-Shirazi, and Nuno Vasconcelos. Cost-sensitive support vector machines. Neurocomputing, 343:50–64, 2019. 15

    2019

  9. [12]

    Integrated parallel k-nearest neighbor algorithm

    Rashmi Agrawal. Integrated parallel k-nearest neighbor algorithm. In Smart Intelligent Computing and Applications, pages 479–486. Springer, 2019

    2019

  10. [13]

    The conformal bootstrap: Theory, numerical tech- niques, and applications

    David Poland, Slava Rychkov, and Alessandro Vichi. The conformal bootstrap: Theory, numerical tech- niques, and applications. Reviews of Modern Physics , 91(1):015002, 2019

    2019

  11. [14]

    Learning a synaptic learning rule

    Yoshua Bengio, Samy Bengio, and Jocelyn Cloutier. Learning a synaptic learning rule . Universit´ e de Montr´ eal, D´ epartement d’informatique et de recherche . . . , 1990

    1990

  12. [15]

    Long short-term memory.Neural computation, 9(8):1735–1780, 1997

    Sepp Hochreiter and J¨ urgen Schmidhuber. Long short-term memory.Neural computation, 9(8):1735–1780, 1997

    1997

  13. [16]

    Deep online learning via meta-learning: Continual adaptation for model-based rl

    Anusha Nagabandi, Chelsea Finn, and Sergey Levine. Deep online learning via meta-learning: Continual adaptation for model-based rl. arXiv preprint arXiv:1812.07671 , 2018

    2018 arXiv

  14. [17]

    Meta- learning with memory-augmented neural networks

    Adam Santoro, Sergey Bartunov, Matthew Botvinick, Daan Wierstra, and Timothy Lillicrap. Meta- learning with memory-augmented neural networks. In International conference on machine learning , pages 1842–1850, 2016

    2016

  15. [19]

    Matching networks for one shot learning

    Oriol Vinyals, Charles Blundell, Timothy Lillicrap, Daan Wierstra, et al. Matching networks for one shot learning. In Advances in neural information processing systems , pages 3630–3638, 2016

    2016

  16. [21]

    How to train your maml

    Antreas Antoniou, Harrison Edwards, and Amos Storkey. How to train your maml. arXiv preprint arXiv:1810.09502, 2018

    2018 arXiv

  17. [22]

    Probabilistic model-agnostic meta-learning

    Chelsea Finn, Kelvin Xu, and Sergey Levine. Probabilistic model-agnostic meta-learning. In Advances in Neural Information Processing Systems, pages 9516–9527, 2018

    2018

  18. [23]

    Prototypical networks for few-shot learning

    Jake Snell, Kevin Swersky, and Richard Zemel. Prototypical networks for few-shot learning. In Advances in Neural Information Processing Systems , pages 4077–4087, 2017

    2017

  19. [24]

    Hierarchical meta learning

    Yingtian Zou and Jiashi Feng. Hierarchical meta learning. arXiv preprint arXiv:1904.09081 , 2019

    1904 arXiv

  20. [25]

    Learning to compare: Relation network for few-shot learning

    Flood Sung, Yongxin Yang, Li Zhang, Tao Xiang, Philip HS Torr, and Timothy M Hospedales. Learning to compare: Relation network for few-shot learning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 1199–1208, 2018. 16

    2018

  21. [26]

    Edge-labeling graph neural network for few-shot learning

    Jongmin Kim, Taesup Kim, Sungwoong Kim, and Chang D Yoo. Edge-labeling graph neural network for few-shot learning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 11–20, 2019

    2019

  22. [29]

    Learning deep representations of fine-grained visual descriptions

    Scott Reed, Zeynep Akata, Honglak Lee, and Bernt Schiele. Learning deep representations of fine-grained visual descriptions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 49–58, 2016

    2016

  23. [30]

    Improving zero-shot learning by mitigating the hubness problem

    Angeliki Lazaridou Georgiana Dinu and Marco Baroni. Improving zero-shot learning by mitigating the hubness problem. In Proceedings of the 3rd International Conference on Learning Representations (ICLR 2015). workshop track, 2015

    2015

  24. [31]

    Semantic autoencoder for zero-shot learning

    Elyor Kodirov, Tao Xiang, and Shaogang Gong. Semantic autoencoder for zero-shot learning. In Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 3174–3183, 2017

    2017

  25. [32]

    Zero-shot learning and knowledge transfer in music classification and tagging

    Jeong Choi, Jongpil Lee, Jiyoung Park, and Juhan Nam. Zero-shot learning and knowledge transfer in music classification and tagging. arXiv preprint arXiv:1906.08615 , 2019

    1906 arXiv

  26. [33]

    Zero-shot visual imitation

    Deepak Pathak, Parsa Mahmoudieh, Guanghao Luo, Pulkit Agrawal, Dian Chen, Yide Shentu, Evan Shelhamer, Jitendra Malik, Alexei A Efros, and Trevor Darrell. Zero-shot visual imitation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops , p...

    2018

  27. [34]

    Efros, and Trevor Darrell

    Deepak Pathak, Parsa Mahmoudieh, Guanghao Luo, Pulkit Agrawal, Dian Chen, Yide Shentu, Evan Shelhamer, Jitendra Malik, Alexei A. Efros, and Trevor Darrell. Zero-shot visual imitation. InInternational Conference on Learning Representations (ICLR), 2018

    2018

  28. [35]

    Human-level concept learning through probabilistic program induction

    Brenden M Lake, Ruslan Salakhutdinov, and Joshua B Tenenbaum. Human-level concept learning through probabilistic program induction. Science, 350(6266):1332–1338, 2015

    2015

  29. [37]

    Meta-transfer learning for few-shot learning

    Qianru Sun, Yaoyao Liu, Tat-Seng Chua, and Bernt Schiele. Meta-transfer learning for few-shot learning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 403–412, 2019

    2019

  30. [38]

    Atmseer: Increasing transparency and controllability in automated machine learn- ing

    Qianwen Wang, Yao Ming, Zhihua Jin, Qiaomu Shen, Dongyu Liu, Micah J Smith, Kalyan Veeramacha- neni, and Huamin Qu. Atmseer: Increasing transparency and controllability in automated machine learn- ing. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Sys...

    2019

Pith tools

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