Pith. sign in

REVIEW 1 major objections 4 minor 45 references

On the ERM Principle in Meta-Learning

T0 review · 1 major / 4 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 The number of examples per task that meta-ERM needs is governed by a dual Helly number.

desk verdict The main characterization is right; the claimed sampling-equivalence flaw in Theorem 4 does not survive contact with the proof, and the paper deserves a serious referee. read the letter →

arxiv 2411.17898 v1 pith:5AEJKV3V submitted 2024-11-26 stat.ML cs.LG

Yannay Alon , Steve Hanneke , Shay Moran , Uri Shalit This is my paper · ORCID
classification stat.MLcs.LG MSC 68T0568Q32
keywords meta-learningempiricalriskminimizationlearningsurfacedualHellynumbersamplecomplexityVCdimensionfew-shotmeta-learnability
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

Meta-learning succeeds when a learner uses data from many related tasks to output a hypothesis class that adapts to a new task. This paper asks how many tasks $n$ and how many labeled examples per task $m$ a meta-empirical-risk minimizer needs, and answers both questions exactly for families containing finitely many hypothesis classes. The number of tasks must always scale as $1/n$, while the number of examples per task obeys a dichotomy governed by the $\varepsilon$-dual Helly number: either a finite $m$ suffices for error to vanish, or the error cannot decay faster than $1/m$. The paper also shows that without the finite-family restriction, even deciding meta-learnability can be independent of ZFC.

What carries the argument

The load-bearing object is the $\varepsilon$-dual Helly number $m_{\mathcal{H}}(\varepsilon)$, the smallest $m$ such that any set of labeled examples that is realizable by some class in the meta-family but has error $>\varepsilon$ against a class $H$ contains a subset of size at most $m$ that already witnesses non-realizability. Its inverse, $\varepsilon_{\mathcal{H}}(m)$, equals the limiting worst-case error of meta-ERM. The parameter controls when a sample can expose a bad class: if $m_{\mathcal{H}}(0)$ is finite, a bounded sample certifies realizability and meta-ERM makes zero error as tasks accumulate; if it is infinite, the witness can be arbitrarily large, forcing error to decay at best as $1/m$. The $n$-lower-bound uses a game-theoretic value $v(\mathcal{H})$ and a distribution over easy and hard domains.

What would settle it

Take a finite VC meta-hypothesis family for which $m_{\mathcal{H}}(0)$ is known to equal $k$, and run a meta-ERM on a realizable meta-distribution using exactly $m=k$ examples per task; if the expected error does not tend to 0 as $n$ grows, Theorem 4's characterization is wrong.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 4: for any finite VC meta-hypothesis family $\mathcal{H}$, the worst-case expected error of meta-ERM as the number of tasks tends to infinity is exactly $\varepsilon_{\mathcal{H}}(m)$, the optimal error function of the $\varepsilon$-dual Helly number. This yields a complete dichotomy: if $m_{\mathcal{H}}(0)$ is finite, then $\varepsilon_{\mathrm{ERM}}^{\mathrm{exp}}(m)=0$ for every $m\ge m_{\mathcal{H}}(0)$, and more generally $m\ge m_{\mathcal{H}}(\varepsilon)$ iff $\varepsilon_{\mathrm{ERM}}^{\mathrm{exp}}(m)\le \varepsilon$; if $m_{\mathcal{H}}(0)=\infty$, then $\varepsilon_{\mathrm{ERM}}^{\mathrm{exp}}(m)=\Omega(1/m)$. The paper further proves a lower bound $\varepsilon_{\mathrm{ERM}}^{\mathrm{dom}}(n)=\Omega(1/n)$ on the task projection for non-trivial families, matching the upper bound $O(1/n)$ from Theorem 1, and establishes (Theorem 5) that for infinite meta-hypothesis families with VC dimension 1, whether $\mathcal{H}^*$ is meta-learnable is independent of ZFC.

Load-bearing premise

The characterization assumes the meta-hypothesis family contains finitely many hypothesis classes, and the paper itself proves that without this finiteness the learnability of a VC-1 family can be independent of ZFC.

Editorial extensions

If this is right

  • Any meta-ERM achieves the optimal $1/n$ task rate among all proper meta-learning algorithms.
  • For families with finite dual Helly number, zero generalization error is achievable with a constant number of examples per task once enough tasks are seen; the exact threshold is $m_{\mathcal{H}}(0)$;
  • For families with infinite dual Helly number, no meta-ERM can beat a $1/m$ rate in $m$, so only the gap between $1/m$ and $\log m/m$ is left to close.
  • The condition $m_{\mathcal{H}}(\varepsilon)$ characterizes a bounded-example-per-task trade-off for every acceptable error $\varepsilon$, giving a finite certificate of meta-learnability.
  • No simple combinatorial dimension can characterize proper meta-learnability in full generality, since a VC-dimension-1 infinite family can be undecidable.

Reading between the lines

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

  • This suggests a practical diagnostic for few-shot methods: estimate the $\varepsilon$-dual Helly number of the class family induced by pretraining; if it is finite, a small fixed shot count should suffice, and the paper predicts exactly which $m$;
  • The $1/m$ barrier implies that for families with unbounded dual Helly number, few-shot performance gains from additional examples per task are fundamentally slow, which may explain diminishing returns in some meta-learning pipelines.
  • One testable extension is to measure meta-ERM error on finite synthetic families, such as finite sets of halfspace classes, and check whether zero error is reached exactly at the predicted $m_{\mathcal{H}}(0)$;
  • The ZFC-independence result suggests that extending the dichotomy to infinite families would require either a definability constraint on the family or a different notion of algorithm; the paper's future-work suggestions of finite Littlestone dimension and parameterized families are the natural routes.
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

1 major / 4 minor

Summary. This paper studies meta-learning in the realizable, distribution-free setting of Baxter, focusing on proper meta-ERM algorithms that output a hypothesis class from a fixed finite meta-hypothesis family H with bounded VC dimension. It introduces the ERM learning surface epsilon_ERM(n,m), its projections in n and m, and a new combinatorial parameter, the epsilon-dual Helly number m_H(epsilon). Theorem 1 gives a general upper bound O_H(1/n + log m/m); Theorem 3 gives an Omega(1/n) lower bound in the number of tasks; Theorem 4 claims the exact equality epsilon_ERM^exp(m) = epsilon_H(m) for the per-task projection, yielding a dichotomy: either finite m suffices for zero error (when m_H(0) < infinity) or the error is Omega(1/m). Theorem 5 shows that dropping finiteness of H can make learnability undecidable in ZFC.

Significance. The claimed characterization is conceptually clean and would resolve the per-task sample complexity of meta-ERM in the finite-family realizable case, strengthening the earlier sufficient condition of Aliakbarpour et al. to a necessary and sufficient one. The lower-bound constructions (Theorem 3, lower bound of Theorem 4) and the EMX-based undecidability result are convincing, and the paper is explicit about the finite-family assumption and proves that it cannot be dropped. However, the upper-bound proof of the central equality rests on an invalid sampling equivalence, so the main claim is not yet established as written.

major comments (1)
  1. [Appendix A.3 (proof of Theorem 4)] The upper bound is not proven as written because the three-step process is not equivalent to the original sampling S ~ Q(n,m). After drawing D_i ~ Q and S'_i ~ D_i^T, the marginal distribution of a uniformly chosen m-subset of S'_i (without replacement) is not D_i^m; for atomic D_i the two distributions can be far in total variation. All subsequent bounds, particularly the 1/(T choose m) probability of containing a witness used in the third event E3, are computed under this different process, so they do not control the original i.i.d. sampling scheme. Since the equality epsilon_ERM^exp(m) = epsilon_H(m) is the paper's central quantitative claim, this is a load-bearing gap. The lower bound appears sound, but a correct upper-bound argument for the original sampling scheme is needed.
minor comments (4)
  1. [Appendix A.3] Event E3 is misprinted as |{i : L_{S_i}(H) >= 0}| = 0, which is always false; it should be |{i : L_{S_i}(H) > 0}| = 0, meaning no m-sample is non-realizable.
  2. [Appendix A.3] In the third-term paragraph, the condition "L_{S'}(H) = 0 and L_{S'}(H) >= beta*gamma*epsilon" is contradictory; the first equality should be removed so that the condition is L_{S'}(H) >= beta*gamma*epsilon.
  3. [Section 4.2 (Definitions 4.4 and 4.5)] The symbol m_H(epsilon) is used both for a single hypothesis class and for the meta-family (the supremum over H in H); these two uses should be distinguished, for instance by writing m_H(epsilon) for the class and m_{H|H}(epsilon) or an overline for the family.
  4. [Proof of Theorem 4, union bound] The union bound after the three exponential estimates writes the sum over H with L_S(H)=0, but the event being bounded is {sup_{H:L_S(H)=0} L_Q(H) >= epsilon}; the sum should be restricted to classes H with L_Q(H) >= epsilon, otherwise the displayed inequality does not follow.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper proves the equality ε_ERM^exp(m) = ε_H(m) as a theorem between independently defined quantities, rather than defining one in terms of the other.

full rationale

The paper's central quantitative claim, Theorem 4, asserts an equality between the ERM learning-surface projection ε_ERM^exp(m) (Definition 2.3) and the optimal error function ε_H(m) (Definition 4.5), which is derived from the ε dual Helly number m_H(ε) (Definitions 4.4 and 2.5). These objects are defined independently: the learning surface is a worst-case expected error of meta-ERM algorithms over realizable meta-distributions, while the ε dual Helly number is a purely combinatorial parameter about witness subsets of non-realizable example sets. The proof idea in Section 4.2 says 'We show two inequalities,' and Appendix A.3 supplies separate upper- and lower-bound arguments. Thus the equality is a derived theorem, not a definitional identity, a fitted parameter renamed as a prediction, or an ansatz smuggled in through citation. Prior work on dual Helly numbers and non-realizable certificates (Bousquet et al. 2020; Aliakbarpour et al. 2023) is used as external context and benchmark; the paper's new ε-Helly variant is defined in the paper itself, and the connection to the learning surface is established by the paper's own proof, so those citations are not load-bearing self-citations. The EMX independence result of Ben-David et al. 2019, which includes one of the present authors, is invoked only for the side remark Theorem 5 about infinite meta-hypothesis families, not to support the main characterization; hence it does not make the central claim circular. The explicit finiteness assumption on the meta-hypothesis family is a stated scope restriction rather than a circular step. A referee concern that the Appendix A.3 upper bound's three-step resampling process is not equivalent to the original i.i.d. sampling is a correctness or rigor gap in the proof; it is not an instance of the conclusion being assumed as an input, and under the provided circularity rules it does not count as circularity. No parameter fitting, no imported uniqueness theorem, and no renaming of a known result as new structure was found. The derivation is self-contained: the main theorems follow from the paper's definitions and standard PAC/VC arguments, and external results are used only as benchmarks or side results.

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

The central claims rest on restrictive but explicitly stated modeling assumptions: finite meta-hypothesis family, bounded VC dimension, realizable binary classification, and i.i.d. domain sampling. The paper shows the finiteness assumption cannot be dropped in general. No free parameters are fitted; mathematical constants are derived. The new definitions (epsilon dual Helly, optimal error function) are descriptors of the family, not invented physical entities.

assumptions (6)
  • domain assumption The meta-hypothesis family H has finitely many classes (|H| < infinity).
    Theorems 1, 2, 3, 4 and Lemma 4.1 assume finiteness; Theorem 5 shows that without it, meta-learnability can be independent of ZFC. Location: Section 2.3 and Section 4.3.
  • domain assumption Each hypothesis class in H is a VC class with VC(H) <= d for a fixed d.
    Definition 2.1; used in the epsilon-net and uniform convergence bounds in Theorem 1 and Theorem 4 proofs. Also ensures downstream task sample complexity.
  • domain assumption The meta-distribution Q is realizable by H, i.e., inf_{H in H} L_Q(H) = 0.
    RE(H) definition in Section 2.1; all results are in the realizable setting, agnostic case left to future work.
  • domain assumption Binary classification with Y = {0,1}.
    Section 2.1 states: 'For binary classification, the setting in interest in this paper, we set Y = {0,1}'.
  • standard math Standard PAC and uniform-convergence bounds for VC classes are used as black boxes.
    Used in the epsilon-net bound in Theorem 1 and in the agnostic uniform convergence bound in Theorem 4's proof; standard results from Vapnik-Chervonenkis theory.
  • standard math ZFC set theory and the EMX independence theorem of Ben-David et al. [2019].
    Theorem 5 relies on the known result that EMX learnability is independent of ZFC; the paper reduces meta-learnability of H* to EMX learnability of F*.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On the ERM Principle in Meta-Learning." pith.science (2026). https://pith.science/paper/5AEJKV3V

@misc{pith2026241117898,
  author       = {Pith},
  title        = {Pith review of: On the ERM Principle in Meta-Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5AEJKV3V}},
  note         = {Machine review of arXiv:2411.17898}
}
abstract

Classic supervised learning involves algorithms trained on $n$ labeled examples to produce a hypothesis $h \in \mathcal{H}$ aimed at performing well on unseen examples. Meta-learning extends this by training across $n$ tasks, with $m$ examples per task, producing a hypothesis class $\mathcal{H}$ within some meta-class $\mathbb{H}$. This setting applies to many modern problems such as in-context learning, hypernetworks, and learning-to-learn. A common method for evaluating the performance of supervised learning algorithms is through their learning curve, which depicts the expected error as a function of the number of training examples. In meta-learning, the learning curve becomes a two-dimensional learning surface, which evaluates the expected error on unseen domains for varying values of $n$ (number of tasks) and $m$ (number of training examples). Our findings characterize the distribution-free learning surfaces of meta-Empirical Risk Minimizers when either $m$ or $n$ tend to infinity: we show that the number of tasks must increase inversely with the desired error. In contrast, we show that the number of examples exhibits very different behavior: it satisfies a dichotomy where every meta-class conforms to one of the following conditions: (i) either $m$ must grow inversely with the error, or (ii) a \emph{finite} number of examples per task suffices for the error to vanish as $n$ goes to infinity. This finding illustrates and characterizes cases in which a small number of examples per task is sufficient for successful learning. We further refine this for positive values of $\varepsilon$ and identify for each $\varepsilon$ how many examples per task are needed to achieve an error of $\varepsilon$ in the limit as the number of tasks $n$ goes to infinity. We achieve this by developing a necessary and sufficient condition for meta-learnability using a bounded number of examples per domain.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 36 canonical work pages

  1. [1]

    Metalearning with Very Few Samples Per Task

    M. Aliakbarpour, K. Bairaktari, G. Brown, A. Smith, N. Srebro, and J. Ullman. Metalearning with very few samples per task. arXiv preprint arXiv:2312.13978, 2023

    work page Pith review arXiv 2023

  2. [2]

    N. Alon, S. Ben-David, N. Cesa-Bianchi, and D. Haussler. Scale-sensitive dimensions, uniform convergence, and learnability. Journal of the ACM (JACM), 44 0 (4): 0 615--631, 1997

    work page 1997

  3. [3]

    Awaysheh, J

    A. Awaysheh, J. Wilcke, F. Elvinger, L. Rees, W. Fan, and K. L. Zimmerman. Review of medical decision support and machine-learning methods. Veterinary pathology, 56 0 (4): 0 512--525, 2019

    work page 2019

  4. [4]

    Baltru s aitis, C

    T. Baltru s aitis, C. Ahuja, and L.-P. Morency. Multimodal machine learning: A survey and taxonomy. IEEE transactions on pattern analysis and machine intelligence, 41 0 (2): 0 423--443, 2018

    work page 2018

  5. [5]

    J. Baxter. A model of inductive bias learning. Journal of artificial intelligence research, 12: 0 149--198, 2000

    work page 2000

  6. [6]

    Ben-David and R

    S. Ben-David and R. Schuller. Exploiting task relatedness for multiple task learning. In Learning Theory and Kernel Machines: 16th Annual Conference on Learning Theory and 7th Kernel Workshop, COLT/Kernel 2003, Washington, DC, USA, August 24-27, 2003. Proceedings, pages 567--580. Springer, 2003

    work page 2003

  7. [7]

    Ben-David, N

    S. Ben-David, N. Cesa-Bianchi, and P. M. Long. Characterizations of learnability for classes of \ O,…, n \ -valued functions. In Proceedings of the fifth annual workshop on Computational learning theory, pages 333--340, 1992

    work page 1992

  8. [8]

    Ben-David, P

    S. Ben-David, P. Hrube s , S. Moran, A. Shpilka, and A. Yehudayoff. Learnability can be undecidable. Nature Machine Intelligence, 1 0 (1): 0 44--48, 2019

    work page 2019

Show all 45 references
  1. [9]

    G. M. Benedek and A. Itai. Nonuniform learnability. Journal of Computer and System Sciences, 48 0 (2): 0 311--323, 1994

    1994

  2. [10]

    W. Bion. Learning from experience. Taylor & Francis, 2023

    2023

  3. [11]

    Blumer, A

    A. Blumer, A. Ehrenfeucht, D. Haussler, and M. K. Warmuth. Learnability and the vapnik-chervonenkis dimension. Journal of the ACM (JACM), 36 0 (4): 0 929--965, 1989

    1989

  4. [12]

    Bousquet, S

    O. Bousquet, S. Hanneke, S. Moran, and N. Zhivotovskiy. Proper learning, helly number, and an optimal svm bound. In Conference on Learning Theory, pages 582--609. PMLR, 2020

    2020

  5. [13]

    Bousquet, S

    O. Bousquet, S. Hanneke, S. Moran, R. Van Handel, and A. Yehudayoff. A theory of universal learning. In Proceedings of the 53rd Annual ACM SIGACT Symposium on Theory of Computing, pages 532--541, 2021

    2021

  6. [14]

    Braverman, G

    M. Braverman, G. Kol, S. Moran, and R. R. Saxena. Convex set disjointness, distributed learning of halfspaces, and lp feasibility. arXiv preprint arXiv:1909.03547, 2019

    1909 arXiv

  7. [15]

    Q. Chen, C. Shui, and M. Marchand. Generalization bounds for meta-learning: An information-theoretic analysis. Advances in Neural Information Processing Systems, 34: 0 25878--25890, 2021

    2021

  8. [16]

    Chowdhary and K

    K. Chowdhary and K. Chowdhary. Natural language processing. Fundamentals of artificial intelligence, pages 603--649, 2020

    2020

  9. [17]

    Daniely, S

    A. Daniely, S. Sabato, S. Ben-David, and S. Shalev-Shwartz. Multiclass learnability and the erm principle. In Proceedings of the 24th Annual Conference on Learning Theory, pages 207--232. JMLR Workshop and Conference Proceedings, 2011

    2011

  10. [18]

    Ehrenfeucht, D

    A. Ehrenfeucht, D. Haussler, M. Kearns, and L. Valiant. A general lower bound on the number of examples needed for learning. Information and Computation, 82 0 (3): 0 247--261, 1989

    1989

  11. [19]

    C. Finn, P. Abbeel, and S. Levine. Model-agnostic meta-learning for fast adaptation of deep networks. In D. Precup and Y. W. Teh, editors, Proceedings of the 34th International Conference on Machine Learning, volume 70 of Proceedings of Machine Learning Research, pages 1126--1...

    2017

  12. [20]

    Galanti, L

    T. Galanti, L. Wolf, and T. Hazan. A theoretical framework for deep transfer learning. Information and Inference: A Journal of the IMA, 5 0 (2): 0 159--209, 2016

    2016

  13. [21]

    S. Hanneke. Refined error bounds for several learning algorithms. Journal of Machine Learning Research, 17 0 (135): 0 1--55, 2016

    2016

  14. [22]

    Hanneke and L

    S. Hanneke and L. Yang. Minimax analysis of active learning. J. Mach. Learn. Res., 16 0 (1): 0 3487--3602, 2015

    2015

  15. [23]

    Hospedales, A

    T. Hospedales, A. Antoniou, P. Micaelli, and A. Storkey. Meta-learning in neural networks: A survey. IEEE transactions on pattern analysis and machine intelligence, 44 0 (9): 0 5149--5169, 2021

    2021

  16. [24]

    Howard and S

    J. Howard and S. Ruder. Universal language model fine-tuning for text classification. arXiv preprint arXiv:1801.06146, 2018

    2018 arXiv

  17. [25]

    A. E. Johnson, M. M. Ghassemi, S. Nemati, K. E. Niehaus, D. A. Clifton, and G. D. Clifford. Machine learning and decision support in critical care. Proceedings of the IEEE, 104 0 (2): 0 444--466, 2016

    2016

  18. [26]

    S. T. Jose and O. Simeone. An information-theoretic analysis of the impact of task similarity on meta-learning. In 2021 IEEE International Symposium on Information Theory (ISIT), pages 1534--1539. IEEE, 2021

    2021

  19. [27]

    Jumper, R

    J. Jumper, R. Evans, A. Pritzel, T. Green, M. Figurnov, O. Ronneberger, K. Tunyasuvunakool, R. Bates, A. Z \' dek, A. Potapenko, et al. Highly accurate protein structure prediction with alphafold. nature, 596 0 (7873): 0 583--589, 2021

    2021

  20. [28]

    D. Kane, R. Livni, S. Moran, and A. Yehudayoff. On communication complexity of classification problems. In Conference on Learning Theory, pages 1903--1943. PMLR, 2019

    1903

  21. [29]

    Khurana, A

    D. Khurana, A. Koli, K. Khatter, and S. Singh. Natural language processing: state of the art, current trends and challenges. Multimedia tools and applications, 82 0 (3): 0 3713--3744, 2023

    2023

  22. [30]

    Konobeev, I

    M. Konobeev, I. Kuzborskij, and C. Szepesv \'a ri. A distribution-dependent analysis of meta learning. In International Conference on Machine Learning, pages 5697--5706. PMLR, 2021

    2021

  23. [31]

    Lucas, M

    J. Lucas, M. Ren, I. Kameni, T. Pitassi, and R. Zemel. Theoretical bounds on estimation error for meta-learning. arXiv preprint arXiv:2010.07140, 2020

    2010 arXiv

  24. [32]

    M. H. Mahmud. On universal transfer learning. Theoretical Computer Science, 410 0 (19): 0 1826--1846, 2009

    2009

  25. [34]

    Nichol, J

    A. Nichol, J. Achiam, and J. Schulman. On first-order meta-learning algorithms. arXiv preprint arXiv:1803.02999, 2018

    2018 arXiv

  26. [35]

    Rezazadeh

    A. Rezazadeh. A unified view on PAC - B ayes bounds for meta-learning. In K. Chaudhuri, S. Jegelka, L. Song, C. Szepesvari, G. Niu, and S. Sabato, editors, Proceedings of the 39th International Conference on Machine Learning, volume 162 of Proceedings of Machine Learning Resea...

    2022

  27. [36]

    S. Ruder. An overview of multi-task learning in deep neural networks. arXiv preprint arXiv:1706.05098, 2017

    2017 arXiv

  28. [37]

    Shalev-Shwartz, O

    S. Shalev-Shwartz, O. Shamir, N. Srebro, and K. Sridharan. Learnability, stability and uniform convergence. The Journal of Machine Learning Research, 11: 0 2635--2670, 2010

    2010

  29. [38]

    M. K. Titsias, F. J. Ruiz, S. Nikoloutsopoulos, and A. Galashov. Information theoretic meta learning with gaussian processes. In Uncertainty in Artificial Intelligence, pages 1597--1606. PMLR, 2021

    2021

  30. [39]

    Torfi, R

    A. Torfi, R. A. Shirvani, Y. Keneshloo, N. Tavaf, and E. A. Fox. Natural language processing advancements by deep learning: A survey. arXiv preprint arXiv:2003.01200, 2020

    2003 arXiv

  31. [40]

    Tripuraneni, M

    N. Tripuraneni, M. Jordan, and C. Jin. On the theory of transfer learning: The importance of task diversity. Advances in neural information processing systems, 33: 0 7852--7862, 2020

    2020

  32. [41]

    Tripuraneni, C

    N. Tripuraneni, C. Jin, and M. Jordan. Provable meta-learning of linear representations. In International Conference on Machine Learning, pages 10434--10443. PMLR, 2021

    2021

  33. [42]

    V. Vapnik. Principles of risk minimization for learning theory. Advances in neural information processing systems, 4, 1991

    1991

  34. [43]

    Vapnik and A

    V. Vapnik and A. Chervonenkis. Theory of pattern recognition. 1974

    1974

  35. [44]

    X. Wang, G. Chen, G. Qian, P. Gao, X.-Y. Wei, Y. Wang, Y. Tian, and W. Gao. Large-scale multi-modal pre-trained models: A comprehensive survey. Machine Intelligence Research, 20 0 (4): 0 447--482, 2023

    2023

  36. [45]

    Y.-X. Wang, J. Lei, and S. E. Fienberg. Learning with differential privacy: Stability, learnability and the sufficiency and necessity of erm principle. Journal of Machine Learning Research, 17 0 (183): 0 1--40, 2016

    2016

  37. [46]

    L. Yang, S. Hanneke, and J. Carbonell. Identifiability of priors from bounded sample sizes with applications to transfer learning. In Proceedings of the 24th Annual Conference on Learning Theory, pages 789--806. JMLR Workshop and Conference Proceedings, 2011

    2011

Pith tools

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