Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Hyperbolic Dual Feature Augmentation for Open-Environment

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that hyperbolic feature augmentation can be extended from closed-set to open-environment learning by synthesizing features for unseen classes as well as seen classes, and that this dual augmentation improves…

desk verdict Useful empirical extension of HFA with consistent gains across five tasks, but the central upper-bound proof is algebraically wrong and the infinite-augmentation claim is unsupported as written. read the letter →

arxiv 2506.08906 v1 pith:ZX5UZQMQ submitted 2025-06-10 cs.CV

Peilin Yu , Yuwei Wu , Zhi Gao , Xiaomeng Fan , Shuo Yang , Yunde Jia This is my paper
classification cs.CV
keywords featureaugmentationhyperbolicspaceopen-environmentlearningneuralordinarydifferentialequationsmeta-learningzero-shotclass-incrementalfew-shot
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 is trying to establish that feature augmentation in hyperbolic space should generate features for unseen classes, not just seen ones, when the model must operate in an open environment where new classes appear over time. It argues that seen-only augmentation lets seen classes over-dominate the embedding space, while dual augmentation reserves room for future classes and sharpens class boundaries. To make the synthetic unseen classes plausible, the method meta-learns neural ODEs that estimate wrapped normal distributions for seen classes, then synthesizes unseen-class distributions by perturbing differences between pairs of seen class distributions. A hierarchy-preserving regularizer keeps synthesized unseen prototypes closer to the center of the Poincaré ball, reflecting their higher uncertainty. The paper also derives an upper bound for the infinite-augmentation cross-entropy loss, so the classifier can be trained with infinitely many augmented features without explicitly sampling them.

What carries the argument

The load-bearing objects are the wrapped normal distribution on the Poincaré ball, whose curvature c, mean μ_i, and covariance factor L_i are estimated by neural ODEs F1, F2, and F3; the unseen-class synthesis via ODE-evolved perturbations δ_p, δ_μ, and δ_L applied to the differences between pairs of seen-class parameters (Eqs. 13-15); and the hierarchy-preserving regularizer R_Hier in Eq. (16), which uses the hyperbolic distance d_c(p_k,0) as an uncertainty measure and pushes synthesized unseen prototypes toward the origin while preventing collapse. The argument is carried by Proposition 1, Eq. (21), an upper bound on the infinite-sample cross-entropy loss obtained through Jensen's inequality, a comparison between hyperbolic distance and Euclidean norm under scale constraints, and the moment-generating function of a Gaussian in the tangent space. Together these pieces convert dual augmentation into a closed-form training objective that avoids explicit hyperbolic sampling.

What would settle it

Take a few-shot open-set or zero-shot task, remove the learned perturbation ODEs, and instead synthesize unseen-class distributions from random directions with matched covariance scale; if the random-direction version matches the reported accuracy and AUROC, the learned perturbation mechanism is not what drives the gains.

Watch

Extended reading notes

Core claim

The paper's central claim is that augmenting features for both seen and unseen classes in hyperbolic space, rather than seen classes only, improves open-environment generalization in class-incremental learning, few-shot open-set recognition, few-shot learning, zero-shot learning, and general image classification. On its own terms, HDFA estimates the seen-class distributions with neural ODEs, synthesizes m = n(n-1)/2 unseen-class distributions by evolving perturbations of the differences between pairs of seen-class parameters through additional ODEs, and regularizes the hyperbolic distance to the origin so that synthesized unseen prototypes sit closer to the center. The technical centerpiece is Proposition 1, which states that Eq. (21) is an upper bound on the expected cross-entropy loss over infinitely many samples drawn from wrapped normal distributions, permitting training without explicit exponential-map sampling. Across Tables 1-10, the method reports higher accuracy or AUROC than the seen-only hyperbolic augmentation baseline HFA in the compared settings.

Load-bearing premise

The load-bearing premise is that unseen test classes are well approximated by synthetic classes created by perturbing differences between pairs of seen class distributions, without using any semantic attributes or external knowledge about what those unseen classes actually are.

Editorial extensions

If this is right

  • Open-environment hyperbolic models can be trained to anticipate classes they have never seen, because synthetic unseen-class features are generated during training from pairwise seen-class differences.
  • The infinite-augmentation upper bound removes the need to sample large batches of hyperbolic features, reducing the computational cost of augmentation-based hyperbolic training.
  • Dual augmentation improves not only open-set detection but also closed-set classification in data-scarce settings, as reported on few-shot and general image classification tasks.
  • The hierarchy regularizer provides a geometric prior for where synthesized classes should sit: unseen classes are placed closer to the center of the Poincaré ball, matching their higher uncertainty.
  • If the claim holds, hyperbolic algorithms can be extended beyond closed-set assumptions to settings where the class set changes over time, without storing samples of future classes.

Reading between the lines

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

  • If the upper-bound derivation is tight in practice, the same Jensen-plus-moment-generating argument could be adapted to other Riemannian manifolds with known distance functions, yielding closed-form infinite augmentation for mixed-curvature or product manifolds.
  • The paper synthesizes unseen classes from pairwise seen-class differences while ignoring semantic attributes; a testable extension would condition the perturbation ODEs on attribute vectors or language descriptions, which could strengthen zero-shot transfer.
  • The central substitution assumption could be probed directly: train HDFA on a subset of classes, then measure how often classifiers for the synthesized unseen classes align with classifiers trained on real held-out classes in feature space.
  • The hierarchy prior's effect likely depends on the chosen curvature; tuning curvature per task could change whether pushing unseen prototypes toward the origin helps or hurts classification.
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 / 5 minor

Summary. The paper proposes Hyperbolic Dual Feature Augmentation (HDFA), a method for open-environment recognition that augments features for both seen and synthesized unseen classes in the Poincaré ball. The method meta-learns neural ODEs to estimate wrapped normal distributions of seen classes, generates auxiliary unseen class distributions by perturbing differences between pairs of seen class distribution parameters, adds a hierarchy-preserving regularizer that places synthetic unseen classes closer to the origin, and derives an upper bound L̄∞ for the infinite-augmentation cross-entropy loss so that a hyperbolic distance classifier can be trained without explicitly sampling augmented features. The method is evaluated on class-incremental learning, few-shot open-set recognition, few-shot learning, zero-shot learning, and general image classification, with reported gains over the seen-only hyperbolic augmentation baseline HFA.

Significance. If the theoretical result held, the upper-bound training would be a useful and efficient way to approximate infinite feature augmentation in hyperbolic space, and the dual augmentation idea could improve open-environment generalization. The strengths of the paper are its broad experimental scope, consistent improvements over HFA across five tasks, ablations that isolate the ODE, infinite-augmentation, and dual-augmentation components, and efficiency comparisons. However, the central theoretical result, Proposition 1, is not established as written: the proof contains an algebraic error in the transition from Eq. (22b) to Eq. (22c), unspecified scale constraints, and an index inconsistency in the displayed bound Eq. (21). In addition, the synthetic unseen class construction is never connected to real test-time unseen classes, which weakens the open-environment and zero-shot claims. These issues are significant but appear to be locally fixable, so the paper merits a major revision rather than rejection.

major comments (4)
  1. [4.5, Eq. (22c)] The transition from Eq. (22b) to Eq. (22c) is algebraically incorrect. Starting from Eq. (22b), the exponent is ||s_{j,i} - w_j||^2 - ||s_{j,i} - w_{j'}||^2. Expanding this gives 2 s_{j,i}^T (w_{j'} - w_j) + ||w_j||^2 - ||w_{j'}||^2. Under the approximation s_{j,i} ≈ p̂_j + v̂_{j,i}, this becomes 2 (p̂_j + v̂_{j,i})^T (w_{j'} - w_j) + ||w_j||^2 - ||w_{j'}||^2. Eq. (22c) instead writes (p̂_j + v̂_{j,i})^T (w_{j'} - w_j) + ||w_{j'}||^2 - ||w_j||^2, which misses the factor of 2 and reverses the sign of the norm term. This is not a scale-constraint issue; the two expressions are algebraically different, including in the small-norm limit. Consequently Eq. (22d) and the final bound Eq. (21) do not follow from the preceding line, and the claimed upper bound on the infinite-augmentation loss is not established.
  2. [4.5, Eq. (21)] Eq. (21) is ill-defined as printed. The outer sum ranges over the class index j, but the distribution mean appears as μ_i, with i not defined in that formula; the same undefined μ_i is used in the numerator and in every term of ξ. The derivation in Eqs. (22c)-(22d) uses μ_j and Σ_j for the class being augmented, so the intended formula almost certainly should have μ_j throughout. The index inconsistency must be fixed in both the numerator and the denominator before the bound can be evaluated.
  3. [4.5, proof of Eq. (22b)] The passage from Eq. (22a) to Eq. (22b) relies on the inequality d_c(s_{j,i}, w_j) - d_c(s_{j,i}, w_{j'}) ≤ ||s_{j,i} - w_j||^2 - ||s_{j,i} - w_{j'}||^2, which the authors state holds only when ||s - w_j||^2 < ||s - w_{j'}||^2 below a curvature-dependent threshold and when additional 'scale constraints on c, ||v̂_{j,i}||, and ||p̂_j||' are imposed. These constraints are never stated precisely, and the estimated parameters c, μ_j, L_j from Eqs. (8)-(9) are not optimized subject to them. The manuscript therefore does not demonstrate that the distributions used in training satisfy the hypotheses of the bound. The authors should state the constraints explicitly and either enforce them during training or provide empirical verification that the learned distributions satisfy them.
  4. [4.3 and 5.4] The synthetic unseen classes constructed in Eqs. (13)-(15) are obtained solely from differences between pairs of seen class distribution parameters. In the zero-shot learning experiments, the true test-time unseen classes are defined by semantic attributes, and these attributes are not used anywhere in the synthesis process. The paper provides no argument or experiment showing that the n(n-1)/2 auxiliary classes approximate actual unseen classes. The observed gains over HFA could therefore reflect a generic regularizing effect of additional classes rather than a faithful model of unseen classes. A control experiment that perturbs seen distributions randomly with the same ODE module, or a quantitative comparison of synthetic and real unseen distribution parameters, would be needed to support the open-environment and zero-shot claims; otherwise the claims should be tempered.
minor comments (5)
  1. [3, Eq. (3)] The definition of the Poincaré ball writes the condition as -c||x|| < 1; it should be -c||x||^2 < 1, since with c = -1 the usual unit ball is ||x|| < 1.
  2. [5.1 and Table 1] The text describes the CIFAR-100 setting C100-B50-S20, while Table 1 labels the corresponding column as C100-B40-S20; these should be unified.
  3. [5.2] The hierarchy regularizer in Eq. (23) includes the hyperparameter β, but the few-shot open-set recognition setup only reports the value of γ. Please state whether β is used in this task and, if so, its value.
  4. [4.5, proof of Proposition 1] The final displayed equality line of the proof contains corrupted notation including 'w_jt' and unbalanced parentheses; the proof must be retyped in full so that the derivation can be checked.
  5. [5.8, Figures 6 and 7] The visualizations show that augmented unseen-class features are closer to the origin, but this property is explicitly enforced by the regularizer in Eq. (16) through the term 2 d_c(p_k,0) - d_c(p_i,0) - d_c(p_j,0). The text should describe these figures as verifying that the regularizer is doing what it was designed to do, not as independent evidence for the uncertainty hypothesis.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor self-definitional validation in hierarchy regularizer; central upper-bound derivation and experiments are otherwise self-contained.

  1. self definitional [Eq. (16) in Section 4.4 and Figure 6 caption in Section 5.8]
    "RHier = 2 ∗ dc(pk, 0) − dc(pi, 0) − dc(pj , 0)+ max (0, γ (dc(pi, 0) + dc(pj , 0)) − dc(pk, 0)) ... The augmented features of unseen classes are closer to the center of the Poincaré ball with higher uncertainty."

    The first term of RHier is minimized exactly when dc(pk,0) is smaller than the average of the seen-class distances, and the max term only prevents collapse to the origin. Thus the property shown in Figure 6 is enforced by the regularizer, not discovered from data. The paper then presents this enforced proximity in Section 5.8 as evidence that the regularizer preserves latent hierarchies and that unseen classes have higher uncertainty, which is a self-confirmation of the design choice. This step is minor: it does not bear on Proposition 1 or on the reported comparisons against external baselines.

full rationale

The central claimed derivation, Proposition 1, is an algebraic upper bound on the infinite-augmentation cross-entropy loss; it uses Jensen’s inequality, a monotonicity condition, and the moment-generating function of the fitted wrapped normal distribution. Even if the proof contains algebraic/indexing errors, that is a correctness problem, not circularity, because the bound is not identical to its inputs by definition. The distribution parameters are fitted to training data, and the classifier is evaluated on held-out test classes, so no fitted parameter is renamed as a prediction. The method builds on HFA (Gao et al., 2022c), but that citation supplies the hyperbolic augmentation recipe and baseline rather than a uniqueness theorem or an unverified premise. The only genuinely circular element is the hierarchy regularizer: RHier explicitly imposes that synthesized unseen-class prototypes lie closer to the origin, and the visualization in Figure 6 then reports this imposed property as an observed confirmation of higher uncertainty. This is a designed constraint rather than a discovered result, but it is a minor validation step and does not undermine the main training objective or the comparative experiments. Overall circularity score: 2.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The method rests on a small set of hand-chosen hyperparameters (beta, gamma, initial curvature) and three paper-specific assumptions: wrapped-normal validity, pairwise interpolated unseen classes, and unproved inequalities in Prop. 1. No new physical entities are introduced.

free parameters (5)
  • initial curvature c0 = -1 (CIL), -0.1 (FSOR), -0.3 (FSL), -0.001 (ZSL)
    Chosen per task, then evolved by the neural ODE; the final curvature is effectively fitted to the data and is not derived from first principles.
  • hierarchy regularizer gamma = 0.1 (searched in [0.01, 0.25])
    Controls how close synthetic unseen prototypes sit to the origin; this directly shapes the placement of all unseen classes.
  • hierarchy regularizer beta = 10
    Weight of R_Hier in the meta-objective Eq. (23); no sensitivity study is reported.
  • CIL augmentation and distillation trade-offs = both 10
    Weights for the augmentation loss and distillation loss in the third CIL training stage; hand-set and not analyzed.
  • ODE integration horizon T
    The number of ODE solve steps is not reported; the estimated distributions and all downstream losses depend on this choice.
assumptions (4)
  • domain assumption Real image feature distributions in the evaluated tasks are well approximated by wrapped normal distributions in a single Poincare ball of constant curvature.
    Used throughout Sect. 4.2-4.5; the conclusion explicitly admits assuming 'a uniform hierarchical structure and a single hyperbolic space for each task'.
  • ad hoc to paper Synthetic unseen classes can be generated by applying learned perturbations to differences between seen class distribution parameters (Eqs. 13-15).
    No evidence connects these interpolated distributions to actual test-time unseen classes; this is the load-bearing modeling choice.
  • ad hoc to paper The inequalities in Prop. 1, including the Jensen step, the Euclidean distance replacement, and the scale constraints, hold for the estimated distributions.
    Eq. (22b) requires squared distances below a curvature-dependent threshold; the paper does not prove this and instead mentions constraints in prose.
  • ad hoc to paper Classes closer to the origin of the Poincare ball have higher uncertainty, so unseen classes should be placed closer to the origin (Eq. 16).
    The regularizer encodes this ordering, so it cannot serve as independent evidence for it.
invented entities (1)
  • Synthetic unseen classes (n(n-1)/2 auxiliary classes)
    purpose: Feature placeholders that reserve embedding space for future or open-set classes during training.
    Their distributions are generated from seen-class pairs and a regularizer; there is no external handle showing they correspond to real unseen categories.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hyperbolic Dual Feature Augmentation for Open-Environment." pith.science (2026). https://pith.science/paper/ZX5UZQMQ

@misc{pith2026250608906,
  author       = {Pith},
  title        = {Pith review of: Hyperbolic Dual Feature Augmentation for Open-Environment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZX5UZQMQ}},
  note         = {Machine review of arXiv:2506.08906}
}
read the original abstract

Feature augmentation generates novel samples in the feature space, providing an effective way to enhance the generalization ability of learning algorithms with hyperbolic geometry. Most hyperbolic feature augmentation is confined to closed-environment, assuming the number of classes is fixed (\emph{i.e.}, seen classes) and generating features only for these classes. In this paper, we propose a hyperbolic dual feature augmentation method for open-environment, which augments features for both seen and unseen classes in the hyperbolic space. To obtain a more precise approximation of the real data distribution for efficient training, (1) we adopt a neural ordinary differential equation module, enhanced by meta-learning, estimating the feature distributions of both seen and unseen classes; (2) we then introduce a regularizer to preserve the latent hierarchical structures of data in the hyperbolic space; (3) we also derive an upper bound for the hyperbolic dual augmentation loss, allowing us to train a hyperbolic model using infinite augmentations for seen and unseen classes. Extensive experiments on five open-environment tasks: class-incremental learning, few-shot open-set recognition, few-shot learning, zero-shot learning, and general image classification, demonstrate that our method effectively enhances the performance of hyperbolic algorithms in open-environment.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Modality Alignment across Trees on Heterogeneous Hyperbolic Manifolds

    cs.CV 2025-10 reject novelty 6.0 of 10

    A VLM method aligns hierarchical image and text feature trees across hyperbolic manifolds of different curvatures via an intermediate manifold, but the theoretical justification is flawed.

Reference graph

Works this paper leans on

131 extracted references · 78 canonical work pages · cited by 1 Pith paper

  1. [1]

    Absil PA, Mahony R, Sepulchre R (2008) Optimization algorithms on matrix manifolds

    2008

  2. [2]

    hairer, s

    Alexander R (1990) Solving ordinary differential equations i: Nonstiff problems (e. hairer, s. p. norsett, and g. wanner). SIAM Review 32(3):485--486

    1990

  3. [3]

    In: European Conference on Computer Vision (ECCV), pp 139--154

    Aljundi R, Babiloni F, Elhoseiny M, et al (2018) Memory aware synapses: Learning what (not) to forget. In: European Conference on Computer Vision (ECCV), pp 139--154

    2018

  4. [4]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 4179--4188

    Anvekar T, Bazazian D (2023) Gpr-net: Geometric prototypical network for point cloud few-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 4179--4188

    2023

  5. [5]

    In: International Conference on Machine Learning (ICML), PMLR, pp 214--223

    Arjovsky M, Chintala S, Bottou L (2017) Wasserstein generative adversarial networks. In: International Conference on Machine Learning (ICML), PMLR, pp 214--223

    2017

  6. [6]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 4453--4462

    Atigh MG, Schoep J, Acar E, et al (2022) Hyperbolic image segmentation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 4453--4462

    2022

  7. [7]

    In: Advances in neural information processing systems (NeurIPS), pp 20755--20765

    Baik S, Choi M, Choi J, et al (2020 a ) Meta-learning with adaptive hyperparameters. In: Advances in neural information processing systems (NeurIPS), pp 20755--20765

    2020

  8. [8]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 2379--2387

    Baik S, Hong S, Lee KM (2020 b ) Learning to forget for meta-learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 2379--2387

    2020

Show all 131 references
  1. [9]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 9465--9474

    Baik S, Choi J, Kim H, et al (2021) Meta-learning with task-adaptive loss function for few-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 9465--9474

    2021

  2. [10]

    In: International Conference on Learning Representations (ICLR)

    Bdeir A, Schwethelm K, Landwehr N (2024) Fully hyperbolic convolutional neural networks for computer vision. In: International Conference on Learning Representations (ICLR)

    2024

  3. [11]

    In: International Conference on Machine Learning (ICML), pp 552--560

    Bengio Y, Mesnil G, Dauphin Y, et al (2013) Better mixing via deep representations. In: International Conference on Machine Learning (ICML), pp 552--560

    2013

  4. [12]

    Flavors of geometry 31:2

    Cannon JW, Floyd WJ, Kenyon R, et al (1997) Hyperbolic geometry. Flavors of geometry 31:2

    1997

  5. [13]

    In: European Conference on Computer Vision (ECCV), pp 233--248

    Castro FM, Mar \' n-Jim \'e nez MJ, Guil N, et al (2018) End-to-end incremental learning. In: European Conference on Computer Vision (ECCV), pp 233--248

    2018

  6. [14]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI) 45(3):2879--2896

    Chadebec C, Thibeau-Sutre E, Burgos N, et al (2022) Data augmentation in high dimensional low sample size setting using a geometry-based variational autoencoder. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI) 45(3):2879--2896

    2022

  7. [15]

    In: International Conference on Learning Representations (ICLR)

    Chaudhry A, Ranzato M, Rohrbach M, et al (2018) Efficient lifelong learning with a-gem. In: International Conference on Learning Representations (ICLR)

    2018

  8. [16]

    IEEE Transactions on Intelligent Transportation Systems (TITS)

    Chen B, Peng W, Cao X, et al (2023) Hyperbolic uncertainty aware semantic segmentation. IEEE Transactions on Intelligent Transportation Systems (TITS)

    2023

  9. [17]

    In: AAAI conference on artificial intelligence (AAAI), pp 10559--10566

    Chen M, Fang Y, Wang X, et al (2020) Diversity transfer network for few-shot learning. In: AAAI conference on artificial intelligence (AAAI), pp 10559--10566

    2020

  10. [18]

    In: Advances in neural information processing systems (NeurIPS)

    Chen RT, Rubanova Y, Bettencourt J, et al (2018) Neural ordinary differential equations. In: Advances in neural information processing systems (NeurIPS)

    2018

  11. [19]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 122--131

    Chen S, Wang W, Xia B, et al (2021 a ) Free: Feature refinement for generalized zero-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 122--131

    2021

  12. [20]

    In: Advances in Neural Information Processing Systems (NeurIPS), pp 16622--16634

    Chen S, Xie G, Liu Y, et al (2021 b ) Hsva: Hierarchical semantic-visual adaptation for zero-shot learning. In: Advances in Neural Information Processing Systems (NeurIPS), pp 16622--16634

    2021

  13. [21]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 8680--8689

    Chen Z, Fu Y, Wang YX, et al (2019 a ) Image deformation meta-networks for one-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 8680--8689

    2019

  14. [22]

    IEEE Transactions on Image Processing (T-IP) 28(9):4594--4605

    Chen Z, Fu Y, Zhang Y, et al (2019 b ) Multi-level semantic feature augmentation for one-shot learning. IEEE Transactions on Image Processing (T-IP) 28(9):4594--4605

    2019

  15. [23]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 248--255

    Deng J, Dong W, Socher R, et al (2009) Imagenet: A large-scale hierarchical image database. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 248--255

    2009

  16. [24]

    IEEE Transactions on Multimedia 25:5763--5774

    Deng S, Yu JG, Wu Z, et al (2022) Learning relative feature displacement for few-shot open-set recognition. IEEE Transactions on Multimedia 25:5763--5774

    2022

  17. [25]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 5138--5146

    Dhar P, Singh RV, Peng KC, et al (2019) Learning without memorizing. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 5138--5146

    2019

  18. [26]

    In: European Conference on Computer Vision (ECCV), pp 86--102

    Douillard A, Cord M, Ollion C, et al (2020) Podnet: Pooled outputs distillation for small-tasks incremental learning. In: European Conference on Computer Vision (ECCV), pp 86--102

    2020

  19. [27]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 10665--10674

    Fang P, Harandi M, Petersson L (2021) Kernel methods in hyperbolic spaces. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 10665--10674

    2021

  20. [28]

    International Journal of Computer Vision (IJCV) 131:2770--2792

    Fang P, Harandi M, Lan Z, et al (2023) Poincar \'e kernels for hyperbolic representations. International Journal of Computer Vision (IJCV) 131:2770--2792

    2023

  21. [29]

    IEEE Transactions on Neural Networks and Learning Systems (TNNLS) 32(6):2506--2520

    Feng L, Zhao C (2020) Transfer increment for generalized zero-shot learning. IEEE Transactions on Neural Networks and Learning Systems (TNNLS) 32(6):2506--2520

    2020

  22. [30]

    In: International Conference on Machine Learning (ICML), pp 1126--1135

    Finn C, Abbeel P, Levine S (2017) Model-agnostic meta-learning for fast adaptation of deep networks. In: International Conference on Machine Learning (ICML), pp 1126--1135

    2017

  23. [31]

    In: European Conference on Computer Vision (ECCV), pp 423--439

    Gao Q, Zhao C, Ghanem B, et al (2022 a ) R-dfcil: Relation-guided representation learning for data-free class incremental learning. In: European Conference on Computer Vision (ECCV), pp 423--439

    2022

  24. [32]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 8691--8700

    Gao Z, Wu Y, Jia Y, et al (2021) Curvature generation in curved spaces for few-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 8691--8700

    2021

  25. [33]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 45:1545--1562

    Gao Z, Wu Y, Harandi M, et al (2022 b ) Curvature-adaptive meta-learning for fast adaptation to manifold data. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 45:1545--1562

    2022

  26. [34]

    In: Advances in neural information processing systems (NeurIPS), pp 34421--34435

    Gao Z, Wu Y, Jia Y, et al (2022 c ) Hyperbolic feature augmentation via distribution estimation and infinite sampling on manifolds. In: Advances in neural information processing systems (NeurIPS), pp 34421--34435

    2022

  27. [35]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 24325--24334

    Gao Z, Xu C, Li F, et al (2023) Exploring data geometry for continual learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 24325--24334

    2023

  28. [36]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 1055--1064

    Gong C, Wang D, Li M, et al (2020) Keepaugment: A simple information-preserving data augmentation approach. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 1055--1064

    2020

  29. [37]

    Communications of the ACM 63(11):139--144

    Goodfellow I, Pouget-Abadie J, Mirza M, et al (2020) Generative adversarial networks. Communications of the ACM 63(11):139--144

    2020

  30. [38]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 16596--16605

    Gudovskiy DA, Rigazio L, Ishizaka S, et al (2021) Autodo: Robust autoaugment for biased data with label noise via scalable probabilistic implicit differentiation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 16596--16605

    2021

  31. [39]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 21--30

    Guo Y, Guo H, Yu SX (2022 a ) Co-sne: Dimensionality reduction and visualization for hyperbolic data. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 21--30

    2022

  32. [40]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 11--20

    Guo Y, Wang X, Chen Y, et al (2022 b ) Clipped hyperbolic classifiers are super-hyperbolic classifiers. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 11--20

    2022

  33. [41]

    Pattern Recognition (PR) 177:151--156

    Hamzaoui M, Chapel L, Pham MT, et al (2024) Hyperbolic prototypical network for few shot remote sensing scene classification. Pattern Recognition (PR) 177:151--156

    2024

  34. [42]

    In: Conference on Empirical Methods in Natural Language Processing (EMNLP), pp 8352--8364

    Han Z, Ding Z, Ma Y, et al (2021) Learning neural ordinary equations for forecasting future links on temporal knowledge graphs. In: Conference on Empirical Methods in Natural Language Processing (EMNLP), pp 8352--8364

    2021

  35. [43]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 770--778

    He K, Zhang X, Ren S, et al (2016) Deep residual learning for image recognition. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 770--778

    2016

  36. [44]

    In: Neural Information Processing SystemsAdvances in neural information processing systems (NeurIPS), pp 6840--6851

    Ho J, Jain A, Abbeel P (2020) Denoising diffusion probabilistic models. In: Neural Information Processing SystemsAdvances in neural information processing systems (NeurIPS), pp 6840--6851

    2020

  37. [45]

    IEEE Transactions on Neural Networks and Learning Systems (TNNLS)

    Hong J, Fang P, Li W, et al (2023 a ) Curved geometric networks for visual anomaly recognition. IEEE Transactions on Neural Networks and Learning Systems (TNNLS)

    2023

  38. [46]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 7873--7883

    Hong J, Hayder Z, Han J, et al (2023 b ) Hyperbolic audio-visual zero-shot learning. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 7873--7883

    2023

  39. [47]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 831--839

    Hou S, Pan X, Loy CC, et al (2019) Learning a unified classifier incrementally via rebalancing. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 831--839

    2019

  40. [48]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 1032--1041

    Hu C, Zhang KY, Yao T, et al (2024) Rethinking generalizable face anti-spoofing via hierarchical prototype-guided distribution refinement in hyperbolic space. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 1032--1041

    2024

  41. [49]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 7171--7180

    Huang S, Ma J, Han G, et al (2022) Task-adaptive negative envision for few-shot open-set recognition. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 7171--7180

    2022

  42. [50]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 4483--4493

    Huynh D, Elhamifar E (2020) Fine-grained generalized zero-shot learning via dense attribute-based attention. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 4483--4493

    2020

  43. [51]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 27621--27630

    Islam K, Zaheer MZ, Mahmood A, et al (2024) Diffusemix: Label-preserving data augmentation with diffusion models. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 27621--27630

    2024

  44. [52]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 12566--12575

    Jeong M, Choi S, Kim C (2021) Few-shot open-set recognition by transformation consistency. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 12566--12575

    2021

  45. [53]

    In: Advances in neural information processing systems (NeurIPS), pp 21655--21667

    Jiang L, Dai B, Wu W, et al (2021) Deceive d: Adaptive pseudo augmentation for gan training with limited data. In: Advances in neural information processing systems (NeurIPS), pp 21655--21667

    2021

  46. [54]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 13300--13308

    Keshari R, Singh R, Vatsa M (2020) Generalized zero-shot learning via over-complete distribution. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 13300--13308

    2020

  47. [55]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 6418--6428

    Khrulkov V, Mirvakhabova L, Ustinova E, et al (2020) Hyperbolic image embeddings. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 6418--6428

    2020

  48. [56]

    Advances in neural information processing systems (NeurIPS) 33:6696--6707

    Kidger P, Morrill J, Foster J, et al (2020) Neural controlled differential equations for irregular time series. Advances in neural information processing systems (NeurIPS) 33:6696--6707

    2020

  49. [57]

    arXiv preprint arXiv:13126114

    Kingma DP (2013) Auto-encoding variational bayes. arXiv preprint arXiv:13126114

    2013

  50. [58]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 9306--9315

    Kong X, Gao Z, Li X, et al (2022) En-compactness: Self-distillation embedding & contrastive generation for generalized zero-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 9306--9315

    2022

  51. [59]

    ://api.semanticscholar.org/CorpusID:18268744

    Krizhevsky A (2009) Learning multiple layers of features from tiny images. ://api.semanticscholar.org/CorpusID:18268744

    2009

  52. [60]

    Communications of the ACM 60:84--90

    Krizhevsky A, Sutskever I, Hinton GE (2012) Imagenet classification with deep convolutional neural networks. Communications of the ACM 60:84--90

    2012

  53. [61]

    IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) pp 2050--2060

    Lazarou M, Avrithis Y, Stathaki T (2021) Tensor feature hallucination for few-shot learning. IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) pp 2050--2060

    2021

  54. [62]

    CS 231N (7):3

    Le Y, Yang X (2015) Tiny imagenet visual recognition challenge. CS 231N (7):3

    2015

  55. [63]

    Lee JM (2006) Riemannian manifolds: an introduction to curvature, vol 176

    2006

  56. [64]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 10657--10665

    Lee K, Maji S, Ravichandran A, et al (2019) Meta-learning with differentiable convex optimization. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 10657--10665

    2019

  57. [65]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 12378--12387

    Li B, Wu F, Lim SN, et al (2020 a ) On feature normalization and data augmentation. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 12378--12387

    2020

  58. [66]

    In: Proceedings of the ACM Web Conference 2022, pp 1528--1537

    Li J, Fu X, Sun Q, et al (2022) Curvature graph generative adversarial networks. In: Proceedings of the ACM Web Conference 2022, pp 1528--1537

    2022

  59. [67]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 13467--13476

    Li K, Zhang Y, Li K, et al (2020 b ) Adversarial feature hallucination networks for few-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 13467--13476

    2020

  60. [68]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 8886--8895

    Li P, Li D, Li W, et al (2021 a ) A simple feature augmentation for domain generalization. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 8886--8895

    2021

  61. [69]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 5208--5217

    Li S, Gong K, Liu CH, et al (2021 b ) Metasaug: Meta semantic augmentation for long-tailed visual recognition. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 5208--5217

    2021

  62. [70]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 11511--11520

    Li S, Xie M, Gong K, et al (2021 c ) Transferable semantic augmentation for domain adaptation. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 11511--11520

    2021

  63. [71]

    In: Advances in neural information processing systems (NeurIPS)

    Lim S, Kim I, Kim T, et al (2019) Fast autoaugment. In: Advances in neural information processing systems (NeurIPS)

    2019

  64. [72]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 8798--8807

    Liu B, Kang H, Li H, et al (2020 a ) Few-shot open-set recognition using meta-learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 8798--8807

    2020

  65. [73]

    In: European Conference on Computer Vision (ECCV), pp 741--756

    Liu J, Song L, Qin Y (2020 b ) Prototype rectification for few-shot learning. In: European Conference on Computer Vision (ECCV), pp 741--756

    2020

  66. [74]

    In: European Conference on Computer Vision (ECCV), pp 699--716

    Liu Y, Parisot S, Slabaugh G, et al (2020 c ) More classifiers, less forgetting: A generic multi-classifier paradigm for incremental learning. In: European Conference on Computer Vision (ECCV), pp 699--716

    2020

  67. [75]

    IEEE Winter Conference on Applications of Computer Vision (WACV) pp 3962--3971

    Luo Q, Wang L, guo Lv J, et al (2021) Few-shot learning via feature hallucination with variational inference. IEEE Winter Conference on Applications of Computer Vision (WACV) pp 3962--3971

    2021

  68. [76]

    In: Advances in neural information processing systems (NeurIPS)

    Mathieu E, Le Lan C, Maddison CJ, et al (2019) Continuous hierarchical representations with poincar \'e variational auto-encoders. In: Advances in neural information processing systems (NeurIPS)

    2019

  69. [77]

    International Journal of Computer Vision (IJCV) pp 1--25

    Mettes P, Ghadimi Atigh M, Keller-Ressel M, et al (2024) Hyperbolic deep learning in computer vision: A survey. International Journal of Computer Vision (IJCV) pp 1--25

    2024

  70. [78]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 12664--12673

    Min S, Yao H, Xie H, et al (2020) Domain-aware visual bias eliminating for generalized zero-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 12664--12673

    2020

  71. [79]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 2082--2090

    Moreira G, Marques M, Costeira JP, et al (2024) Hyperbolic vs euclidean embeddings in few-shot learning: Two sides of the same coin. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 2082--2090

    2024

  72. [80]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI)

    Nag S, Raychaudhuri DS, Paul S, et al (2023) Reconstruction guided meta-learning for few shot open set recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI)

    2023

  73. [81]

    In: IEEE/CVF Winter Conference on Applications of Computer Vision (WCACV), pp 3911--3920

    Petit G, Popescu A, Schindler H, et al (2023) Fetril: Feature translation for exemplar-free class-incremental learning. In: IEEE/CVF Winter Conference on Applications of Computer Vision (WCACV), pp 3911--3920

    2023

  74. [82]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 2001--2010

    Rebuffi SA, Kolesnikov A, Sperl G, et al (2017) icarl: Incremental classifier and representation learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 2001--2010

    2017

  75. [83]

    In: International Conference on Learning Representations (ICLR)

    Ren M, Triantafillou E, Ravi S, et al (2018) Meta-learning for semi-supervised few-shot classification. In: International Conference on Learning Representations (ICLR)

    2018

  76. [84]

    In: Conference on Empirical Methods in Natural Language Processing, pp 9858--9868

    Sawhney R, Thakkar M, Agarwal S, et al (2021) Hypmix: hyperbolic interpolative data augmentation. In: Conference on Empirical Methods in Natural Language Processing, pp 9858--9868

    2021

  77. [85]

    In: Advances in neural information processing systems (NeurIPS)

    Schwartz E, Karlinsky L, Shtok J, et al (2018) Delta-encoder: an effective sample synthesis method for few-shot object recognition. In: Advances in neural information processing systems (NeurIPS)

    2018

  78. [86]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 4136--4145

    Simon C, Koniusz P, Nock R, et al (2020) Adaptive subspaces for few-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 4136--4145

    2020

  79. [87]

    arXiv preprint arXiv:200611328

    Skorokhodov I, Elhoseiny M (2020) Class normalization for (continual)? generalized zero-shot learning. arXiv preprint arXiv:200611328

    2020

  80. [88]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 9374--9384

    Smith J, Hsu YC, Balloch J, et al (2021) Always be dreaming: A new approach for data-free class-incremental learning. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 9374--9384

    2021

  81. [89]

    In: Advances in neural information processing systems (NeurIPS)

    Snell J, Swersky K, Zemel R (2017) Prototypical networks for few-shot learning. In: Advances in neural information processing systems (NeurIPS)

    2017

  82. [90]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 5419--5428

    van Spengler M, Berkhout E, Mettes P (2023) Poincar \'e resnet. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 5419--5428

    2023

  83. [91]

    Pattern Recognition 151:110400

    Sun LY, Chu WT (2024) Overall positive prototype for few-shot open-set recognition. Pattern Recognition 151:110400

    2024

  84. [92]

    In: Advances in neural information processing systems (NeurIPS)

    Thulasidasan S, Chennupati G, Bilmes JA, et al (2019) On mixup training: Improved calibration and predictive uncertainty for deep neural networks. In: Advances in neural information processing systems (NeurIPS)

    2019

  85. [93]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 7064--7073

    Upchurch P, Gardner J, Pleiss G, et al (2017) Deep feature interpolation for image content changes. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 7064--7073

    2017

  86. [94]

    In: International Conference on Machine Learning (ICML)

    Verma V, Lamb A, Beckham C, et al (2018) Manifold mixup: Better representations by interpolating hidden states. In: International Conference on Machine Learning (ICML)

    2018

  87. [95]

    In: Advances in neural information processing systems (NeurIPS)

    Vinyals O, Blundell C, Lillicrap T, et al (2016) Matching networks for one shot learning. In: Advances in neural information processing systems (NeurIPS)

    2016

  88. [96]

    In: European Conference on Computer Vision (ECCV), pp 70--86

    Vyas MR, Venkateswara H, Panchanathan S (2020) Leveraging seen and unseen semantic relationships for generative zero-shot learning. In: European Conference on Computer Vision (ECCV), pp 70--86

    2020

  89. [97]

    Wah C, Branson S, Welinder P, et al (2011) The caltech-ucsd birds-200-2011 dataset

    2011

  90. [98]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 7507--7516

    Wang H, Pang G, Wang P, et al (2023) Glocal energy-based learning for few-shot open-set recognition. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 7507--7516

    2023

  91. [99]

    In: Advances in neural information processing systems (NeurIPS)

    Wang Y, Pan X, Song S, et al (2019) Implicit semantic data augmentation for deep networks. In: Advances in neural information processing systems (NeurIPS)

    2019

  92. [100]

    IEEE Transactions on Image Processing (TIP) 33:1768--1781

    Wei R, Liu Y, Song J, et al (2024) Exploring hierarchical information in hyperbolic space for self-supervised image hashing. IEEE Transactions on Image Processing (TIP) 33:1768--1781

    2024

  93. [101]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 374--382

    Wu Y, Chen Y, Wang L, et al (2019) Large scale incremental learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 374--382

    2019

  94. [102]

    IEEE transactions on pattern analysis and machine intelligence (TPAMI) 41(9):2251--2265

    Xian Y, Lampert CH, Schiele B, et al (2018) Zero-shot learning—a comprehensive evaluation of the good, the bad and the ugly. IEEE transactions on pattern analysis and machine intelligence (TPAMI) 41(9):2251--2265

    2018

  95. [103]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 9384--9393

    Xie GS, Liu L, Jin X, et al (2019) Attentive region embedding network for zero-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 9384--9393

    2019

  96. [104]

    In: European Conference on Computer Vision (ECCV), pp 562--580

    Xie GS, Liu L, Zhu F, et al (2020) Region graph embedding network for zero-shot learning. In: European Conference on Computer Vision (ECCV), pp 562--580

    2020

  97. [105]

    In: Advances in neural information processing systems (NeurIPS), pp 21969--21980

    Xu W, Xian Y, Wang J, et al (2020) Attribute prototype network for zero-shot learning. In: Advances in neural information processing systems (NeurIPS), pp 21969--21980

    2020

  98. [106]

    Neurocomputing 491:57--66

    Xu Y, Mu L, Ji Z, et al (2022) Meta hyperbolic networks for zero-shot learning. Neurocomputing 491:57--66

    2022

  99. [107]

    In: AAAI Conference on Artificial Intelligence (AAAI), pp 16247--16254

    Xue Y, Dai J, Lu Z, et al (2024) Residual hyperbolic graph convolution networks. In: AAAI Conference on Artificial Intelligence (AAAI), pp 16247--16254

    2024

  100. [108]

    In: ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), pp 3770--3781

    Yang M, Verma H, Zhang DC, et al (2024) Hypformer: Exploring efficient transformer fully in hyperbolic space. In: ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), pp 3770--3781

    2024

  101. [109]

    In: International Conference on Machine Learning (ICML), pp 11887--11897

    Yao H, Huang LK, Zhang L, et al (2021 a ) Improving generalization in meta-learning via task augmentation. In: International Conference on Machine Learning (ICML), pp 11887--11897

    2021

  102. [110]

    In: International Conference on Learning Representations (ICLR)

    Yao H, Zhang L, Finn C (2021 b ) Meta-learning with fewer tasks through task interpolation. In: International Conference on Learning Representations (ICLR)

    2021

  103. [111]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 8808--8817

    Ye HJ, Hu H, Zhan DC, et al (2020) Few-shot learning via embedding adaptation with set-to-set functions. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 8808--8817

    2020

  104. [112]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 6982--6991

    Yu L, Twardowski B, Liu X, et al (2020 a ) Semantic drift compensation for class-incremental learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 6982--6991

    2020

  105. [113]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 14035--14044

    Yu Y, Ji Z, Han J, et al (2020 b ) Episode-based prototype generating network for zero-shot learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 14035--14044

    2020

  106. [114]

    In: IEEE/CVF Winter Conference on Applications of Computer Vision (WCACV), pp 595--605

    Yuan J, Phan MH, Liu L, et al (2024) Fakd: Feature augmented knowledge distillation for semantic segmentation. In: IEEE/CVF Winter Conference on Applications of Computer Vision (WCACV), pp 595--605

    2024

  107. [115]

    In: IEEE/CVF Winter Conference on Applications of Computer Vision (WCACV), pp 1891--1903

    Yue Y, Lin F, Mou G, et al (2024) Understanding hyperbolic metric learning through hard negative sampling. In: IEEE/CVF Winter Conference on Applications of Computer Vision (WCACV), pp 1891--1903

    2024

  108. [116]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 15404--15414

    Yue Z, Wang T, Sun Q, et al (2021) Counterfactual zero-shot and open-set visual recognition. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 15404--15414

    2021

  109. [117]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 6023--6032

    Yun S, Han D, Oh SJ, et al (2019) Cutmix: Regularization strategy to train strong classifiers with localizable features. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 6023--6032

    2019

  110. [118]

    arXiv preprint arXiv:160507146

    Zagoruyko S, Komodakis N (2016) Wide residual networks. arXiv preprint arXiv:160507146

    2016

  111. [119]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 3457--3466

    Zang Y, Huang C, Loy CC (2021) Fasa: Feature augmentation and sampling adaptation for long-tailed instance segmentation. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 3457--3466

    2021

  112. [120]

    In: AAAI conference on artificial intelligence (AAAI), pp 9014--9021

    Zhang B, Li X, Feng S, et al (2022) Metanode: Prototype optimization as a neural ode for few-shot learning. In: AAAI conference on artificial intelligence (AAAI), pp 9014--9021

    2022

  113. [121]

    In: International Conference on Learning Representations (ICLR)

    Zhang H, Cisse M, Dauphin YN, et al (2018) mixup: Beyond empirical risk minimization. In: International Conference on Learning Representations (ICLR)

    2018

  114. [122]

    IEEE Transactions on Neural Networks and Learning Systems (TNNLS) 31(7):2361--2375

    Zhang H, Mao H, Long Y, et al (2019 a ) A probabilistic zero-shot learning method via latent nonnegative prototype synthesis of unseen classes. IEEE Transactions on Neural Networks and Learning Systems (TNNLS) 31(7):2361--2375

    2019

  115. [123]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 2770--2779

    Zhang H, Zhang J, Koniusz P (2019 b ) Few-shot learning via saliency-guided hallucination of samples. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 2770--2779

    2019

  116. [124]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 1685--1694

    Zhang J, Zhao C, Ni B, et al (2019 c ) Variational few-shot learning. In: IEEE/CVF International Conference on Computer Vision (ICCV), pp 1685--1694

    2019

  117. [125]

    Pattern Recognition 136:109235

    Zhang L, Zhou F, Wei W, et al (2023) Meta-hallucinating prototype for few-shot learning promotion. Pattern Recognition 136:109235

    2023

  118. [126]

    In: Advances in neural information processing systems (NeurIPS), pp 49403--49415

    Zheng K, Zhang H, Huang W (2023) Diffkendall: a novel approach for few-shot learning with differentiable kendall's rank correlation. In: Advances in neural information processing systems (NeurIPS), pp 49403--49415

    2023

  119. [127]

    In: Advances in neural information processing systems (NeurIPS), pp 14306--14318

    Zhu F, Cheng Z, Zhang XY, et al (2021 a ) Class-incremental learning via dual augmentation. In: Advances in neural information processing systems (NeurIPS), pp 14306--14318

    2021

  120. [128]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 5871--5880

    Zhu F, Zhang XY, Wang C, et al (2021 b ) Prototype augmentation and self-supervision for incremental learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 5871--5880

    2021

  121. [129]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 45:7477--7493

    Zhu F, Zhang XY, Wang RQ, et al (2022) Learning by seeing more classes. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 45:7477--7493

    2022

  122. [130]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter doi edition editor eid eprint howpublished institution journal key keywords month note number organization pages publisher school series title type url volume year archivePrefix primaryClass adsurl adsnote version label extra.labe...

  123. [131]

    write newline

    " write newline "" before.all 'output.state := FUNCTION add.period duplicate empty 'skip "." * add.blank if FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap dupl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.