Pith. sign in

REVIEW 4 major objections 4 minor 62 references

Partial Domain Adaptation via Importance Sampling-based Shift Correction

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Sampling new labeled data from a constructed distribution, rather than reweighting old source samples, controls target risk in partial domain adaptation through a four-term error bound.

desk verdict A promising sampling-based PDA method with a real O(n²) ETIC speedup, but Proposition 1 proves far less than claimed and the empirical SOTA claims are contradicted by the paper's own tables. read the letter →

arxiv 2507.20191 v1 pith:3ORNOJOE submitted 2025-07-27 cs.LG cs.AI

Cheng-Jun Guo , Chuan-Xian Ren , You-Wei Luo , Xiao-Lin Xu , Hong Yan This is my paper · ORCID
classification cs.LGcs.AI
keywords partialdomainadaptationimportancesamplinggeneralizationerroranalysislabelshiftconditionaloptimaltransportindependencecriterionweighting
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper makes the case that label-shift correction in partial domain adaptation does not have to happen by reweighting source samples. Instead, it proposes to build a new 'sampling domain': draw pairs of same-class source samples, mix them with ratio $\theta$, and use the mixture as fresh labeled data with the target label proportions. The central theoretical claim is that training on this sampling domain controls target risk, because the gap between sampling and target errors is bounded by a label-shift term, a conditional-shift term, and a mixing penalty. A second proposition says that, when model outputs are convex, the sampling-domain model has no larger risk than the source-domain or reweighted model. If these claims hold, importance sampling becomes a principled replacement for importance weighting, with the practical benefit that outlier classes vanish from training and the model can exploit synthetic but labeled data.

What carries the argument

The machinery is the sampling distribution $P_c$: for each class $j$, draw two source samples $X_1,X_2$ from $P_s(X\mid Y=j)$ and form $X=\theta X_1+(1-\theta)X_2$, while setting $p_c(Y=j)=p_t(Y=j)$. Theorem 1 is the identity that carries the argument, decomposing the sampling-to-target error gap into the balanced prediction error, the conditional error gap $\Delta_{CE}$, and the mixing penalty $2\ell(K-1)\sqrt{\theta(1-\theta)}\,C_s$. Proposition 1 then compares $P_c$ with the degenerate cases $\theta=0,1$ under convexity, showing the sampling-domain risk is no larger. The alignment side of the method is ETIC, an entropy-regularized optimal transport independence criterion, whose empirical estimator is accelerated from $\mathcal{O}(n_j^3)$ to $\mathcal{O}(n_j^2)$ per iteration by using the discreteness of the domain variable.

What would settle it

Train a fixed non-convex network on a fixed source domain and compare the empirical risk on the sampled domain $P_c$ with the empirical risk on the weighted-source domain for the same target label proportions; if the sampling-domain risk ever exceeds the weighted-source risk after convergence, the claimed dominance of sampling over reweighting beyond the convex case is falsified. The paper's own Figures 6(c) and 6(d) cover only two real-data checkpoints, so a systematic sweep over tasks and seeds can settle it.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that a model trained on the sampled domain $P_c$ --- built from same-class convex mixtures $X = \theta X_1 + (1 - \theta) X_2$ with $p_c(Y)=p_t(Y)$ --- controls the target risk through the inequality $$|\varepsilon_t(h\circ g) - \varepsilon_c(h\circ g)| \le \|P_c(Y)-P_t(Y)\|_1 \Delta_{BE}(P_c)(\hat Y \| Y) + (K-1)\Delta_{CE}(\hat Y) + 2\ell(K-1)\sqrt{\$\theta$(1-\$\theta$)}C_s.$$ Each term is interpretable: the risk on the sampling domain, the label-distribution gap times the balanced prediction error, the class-conditional gap across shared classes, and a penalty from mixing. Under the additional condition that every component of $h \circ g$ is convex in the input, Proposition 1 gives $\varepsilon_c(h\circ g) \le \varepsilon_{c'}(h\circ g)$, where $P_{c'}$ is the unmixed source sampling domain, so the mixture never hurts and typically helps. The method then estimates the target label proportions with BBSE, samples from $P_c$, and trains by minimizing the classification loss on the sampled data together with an optimal-transport independence criterion (ETIC) for class-conditional alignment. Experiments across Office-31, Office-Home, Image-CLEF and VisDA-2017 are reported as consistent with these bounds, with accuracy gains over reweighting baselines.

Load-bearing premise

The load-bearing premise is that the model's output is convex in the input; the paper itself calls this assumption 'quite strong,' and the neural network used in experiments is not convex.

Editorial extensions

If this is right

  • If Theorem 1 holds, then PDA training can be reframed as risk minimization on a synthetic but labeled domain, so outlier classes drop out of the training set rather than lingering with small weights.
  • The four-term bound makes the mix-ratio $\theta$ a directly tunable trade-off: $\theta$ near 0 or 1 removes the mixing penalty but leaves cluster structure loose, while $\theta$ near 0.5 tightens clusters but adds penalty, predicting the observed U-shaped sensitivity.
  • With the fast ETIC computation, class-conditional alignment costs $\mathcal{O}(n^2)$ per iteration instead of $\mathcal{O}(n^3)$, making the method practical for larger target batches in PDA.
  • The BBSE-based estimate of $p_t(Y)$ means the method does not require known target label proportions; it only needs pseudo-labels from the current model.
  • If Proposition 1 extends beyond convex models, sampling-based correction should dominate importance weighting in accuracy across many PDA benchmarks, not just the four reported datasets.

Reading between the lines

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

  • Editorial inference: the same mixture construction is a natural template for open-set domain adaptation, where the target includes unknown classes; one would need to allocate a mass to an 'unknown' class rather than deleting outliers, a direction the paper names as future work.
  • Editorial inference: because the bound's mixing penalty scales with $\sqrt{\theta(1-\theta)}$ and the source radius $C_s$, normalizing or whitening features before sampling could tighten the bound in high-dimensional feature spaces; the paper does not test this.
  • Editorial inference: if Proposition 1's convexity is genuinely unnecessary, then a direct re-derivation through interpolation or Jensen-type inequalities may yield a weaker but more general dominance statement; the paper's evidence is only two real-data checkpoints in Figures 6(c) and 6(d).
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

4 major / 4 minor

Summary. The paper proposes IS2C for partial domain adaptation. Instead of reweighting source samples, IS2C constructs a sampling domain whose label distribution matches the estimated target label distribution and whose class-conditional features are pairwise convex combinations of source features from the same class. The paper gives Theorem 1, an upper bound on |epsilon_t(h.g)-epsilon_c(h.g)| in terms of label-shift, conditional-shift, and mixture terms, and Proposition 1, which asserts that the risk on the sampling domain is no larger than the risk on the original or reweighted source under a convexity condition. An OT-based conditional alignment loss (ETIC) with a claimed O(n^2) per-iteration cost is proposed, and experiments are reported on Office-31, Office-Home, VisDA-2017, and Image-CLEF.

Significance. If valid, the theory would provide a principled sampling alternative to importance weighting for PDA, with a transparent decomposition of the target-versus-sampling risk gap, and the ETIC speedup would be practically useful. The paper also contains a clear ablation study and sensitivity analysis. However, the main comparison result (Proposition 1) is either vacuous under the paper's own definitions or false under the standard 0-1 argmax risk, and the experimental results do not consistently support superiority over the MOT baseline. The central claim that sampling is theoretically preferable to reweighting is therefore not established.

major comments (4)
  1. [Section III-B, Proposition 1] Under the paper's own definition in Section III, h.g is a K-dimensional probabilistic vector, so each component f_i = (h.g)_i is nonnegative and sum_i f_i = 1. If every f_i is convex, then the constant function 1 = sum_i f_i is convex, and Jensen's inequality must be tight for the sum, forcing each f_i to be affine. Hence the premise of Proposition 1 is compatible only with affine models, for which the risk is unchanged by the mixture operation under the paper's random-prediction risk. The claimed 'strictly smaller risk' for non-linear models does not follow from the stated assumption, so the proposition cannot justify preferring sampling over importance weighting.
  2. [Section III-B, Proposition 1 (0-1 argmax interpretation)] If the error is instead interpreted as the conventional 0-1 argmax error, which is the interpretation used in the experimental visualizations and decision-boundary figures, the asserted inequality is false. For example, take K=2, X in [0,1], P(Y=1)=0.1, P(Y=2)=0.9, X|Y=1 ~ U[0.2,1], X|Y=2 ~ U[0.8,1], and h.g(x)=(1-x,x). For theta=0 or 1, the risk is 0.1 * P(U[0.2,1] > 0.5) = 0.0625. For theta=0.5, X|Y=1 is the average of two independent U[0.2,1] variables and P(X > 0.5) = 0.71875, giving epsilon_c = 0.071875 > epsilon_c'. Thus the proposition cannot be correct under that risk.
  3. [Section III-B, proofs] The proofs of Theorem 1 and Proposition 1 are deferred to a supplementary file that is not included in the arXiv submission. Since Proposition 1 is central to the paper's claim that sampling is at least as good as reweighting, and since that proposition appears to be unsound, the missing proofs are not merely a presentation issue; the derivation of the main theoretical guarantee cannot be checked.
  4. [Section V-B, Table I] The claimed empirical superiority is contradicted by the paper's own numbers. On Office-Home, MOT achieves mean accuracy 80.6% while IS2C achieves 79.2%; on VisDA-2017, MOT achieves 92.4% while IS2C achieves 89.3%; on Image-CLEF, MOT achieves 93.6% while IS2C achieves 93.2%. The sentence immediately after Table I stating that IS2C outperforms comparison methods by at least about 1.0% does not hold against MOT, so the empirical case for IS2C over the strongest baseline is not established.
minor comments (4)
  1. [Section IV-B] The phrase 'addictive cost functions' should be 'additive cost functions'.
  2. [Section IV-C, Algorithm 1] In line 4 of Algorithm 1, the displayed formula p_t(Y) = p_s(Y) * p_t(Y)/p_s(Y) is tautological; please state the BBSE estimation step explicitly as estimating the ratio and then renormalizing.
  3. [Section III-B and Section V-C] The claim that theta close to 0.5 creates an 'overly compact cluster structure' is informal; no quantitative measure of cluster compactness or its effect on the target risk is provided.
  4. [Figure 6] The caption and surrounding text should clarify that the synthetic validation compares the sampling-domain risk with the source-domain risk, and that the source-domain risk is intended to represent epsilon_c' in Proposition 1.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the bound and the sampling comparison are stand-alone theoretical claims, and the definitional equality P_c(Y)=P_t(Y) is an explicit construction assumption, not a fitted prediction.

full rationale

The derivation chain is non-circular. Theorem 1 bounds |ε_t(h∘g)−ε_c(h∘g)| by the label-shift term ||P_c(Y)−P_t(Y)||_1·Δ_BE, the conditional error gap Δ_CE, and a Lipschitz/mixture term. The label-shift term is indeed identically zero under Definition 1, which constructs P_c with p_c(Y=j)=p_t(Y=j); however, this is an explicit definitional assumption of the sampling domain, not a quantity estimated from data and then presented as a prediction. The paper even acknowledges that Theorem 1 itself places no constraint on P_c(Y), and that IS²C is what builds P_c to approximate the target label distribution. The remaining terms of the bound are genuine: Δ_BE is minimized by risk minimization on P_c, Δ_CE is minimized by the ETIC conditional-alignment objective, and the mixture term is a Lipschitz penalty. Proposition 1 is a standalone comparison between two sampling domains under a convexity assumption; whether that assumption is correct or too strong is a mathematical-correctness question, not circularity. The authors themselves flag that the convexity assumption is 'quite strong' and not satisfied by the softmax network, which is a limitation statement rather than a circular derivation. No load-bearing self-citation chain is present: ETIC is taken from an external paper [49], and the definitions of balanced prediction error and conditional error gap are taken from [22]. The experimental comparisons are against external PDA benchmarks, and the risk-on-sampling-domain validation trains on source data and tests on source/sampling data, so no fitted parameter is renamed as a prediction. Overall, the central claims do not reduce to their own inputs.

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

The central method rests on several assumptions beyond standard supervised learning: BBSE must recover target label proportions despite conditional shift; mixing two same-class inputs must preserve the label; Theorem 1 requires an ell-Lipschitz model and Proposition 1 requires convex output components, neither of which holds for the softmax network actually trained; and ETIC's independence characterization is imported from prior work. The main tuned quantities are mu, alpha, and n_c, with no formal selection rule or error bars.

free parameters (4)
  • balance parameter mu = grid {75,100,125,150,175,200,225,250}; final value not individually reported
    Weights the alignment loss in L = L_risk + mu L_align (Eq. 7); tuned per task, no selection rule or error bars given.
  • Beta sampling parameter alpha = 0.2
    Controls the distribution of mix-ratio theta; chosen from {0.1,0.2,0.4,2,4,8} with best at 0.2-0.4 reported in Section V-C.
  • sampling domain size n_c = 2 n_s
    Number of generated labeled samples, set to twice source size in Section IV-C; no sensitivity analysis for n_c is provided.
  • mix-ratio theta (per draw) = sampled from Beta(0.2,0.2)
    The central interpolation parameter in Definition 1; results show best fixed theta near 0.1 or 0.9, while random sampling from Beta(0.2,0.2) is used in training.
assumptions (4)
  • domain assumption BBSE recoverability of target label proportions under label shift.
    Algorithm 1 estimates p_t(Y) by solving the BBSE quadratic program (Section IV-C). BBSE's consistency assumes source and target share class-conditional feature distributions; PDA violates this by construction, so the estimate is biased unless conditional alignment succeeds first.
  • domain assumption Within-class convex interpolation preserves the semantic label.
    Definition 1 assigns label j to any X = theta X1 + (1-theta) X2 with X1,X2 drawn from class j. This assumes class manifolds are convex enough that mixtures remain valid members of the same class.
  • ad hoc to paper Model h.g is ell-Lipschitz (Theorem 1) and each output component is convex (Proposition 1).
    These assumptions appear in Section III-B; the softmax network used in experiments is not convex, and the paper provides only informal experiments for the non-convex case.
  • domain assumption ETIC zero-iff-independence property inherited from prior work.
    The alignment loss relies on T_epsilon(g(X),Z|Y=j)=0 iff conditional independence (Section IV-A), a result proved for the entropy-regularized OT independence criterion in [49], not re-derived here.
invented entities (1)
  • Sampling domain P_c
    purpose: A synthetic labeled distribution with target label proportions and mixed class-conditional source features, used to generate training data and justify the method.
    P_c is constructed by Definition 1 rather than observed; its benefit is only evidenced by benchmark accuracy, which is mixed against MOT, and it has no falsifiable prediction outside the method itself.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Partial Domain Adaptation via Importance Sampling-based Shift Correction." pith.science (2026). https://pith.science/paper/3ORNOJOE

@misc{pith2026250720191,
  author       = {Pith},
  title        = {Pith review of: Partial Domain Adaptation via Importance Sampling-based Shift Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3ORNOJOE}},
  note         = {Machine review of arXiv:2507.20191}
}
abstract

Partial domain adaptation (PDA) is a challenging task in real-world machine learning scenarios. It aims to transfer knowledge from a labeled source domain to a related unlabeled target domain, where the support set of the source label distribution subsumes the target one. Previous PDA works managed to correct the label distribution shift by weighting samples in the source domain. However, the simple reweighing technique cannot explore the latent structure and sufficiently use the labeled data, and then models are prone to over-fitting on the source domain. In this work, we propose a novel importance sampling-based shift correction (IS$^2$C) method, where new labeled data are sampled from a built sampling domain, whose label distribution is supposed to be the same as the target domain, to characterize the latent structure and enhance the generalization ability of the model. We provide theoretical guarantees for IS$^2$C by proving that the generalization error can be sufficiently dominated by IS$^2$C. In particular, by implementing sampling with the mixture distribution, the extent of shift between source and sampling domains can be connected to generalization error, which provides an interpretable way to build IS$^2$C. To improve knowledge transfer, an optimal transport-based independence criterion is proposed for conditional distribution alignment, where the computation of the criterion can be adjusted to reduce the complexity from $\mathcal{O}(n^3)$ to $\mathcal{O}(n^2)$ in realistic PDA scenarios. Extensive experiments on PDA benchmarks validate the theoretical results and demonstrate the effectiveness of our IS$^2$C over existing methods.

Figures

Figures reproduced from arXiv: 2507.20191 by the authors.

Figure 1
Figure 1. Partial domain adaptation, in which the support of source label [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. A simple illustration of the reason for choosing [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. An intuitive illustration of IS2C under Theorem 1. Different colors represent different classes. “◦” represents the source domain, “△” represents the sampling domain, and “+” represents the target domain. The dashed lines represent decision boundaries. (a). The original source domain where exists label shift and conditional shift, and the target domain. (b)-(c). The sampling domains with different mix-ratio and the … view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: (a). Comparison between importance sampling strategies, importance weighting strategies, and mixup augmentation, which are denoted as IS [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Sensitivity analysis on hyper-parameters on Office-Home task Ar [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Model error of h ◦ g over training epochs on the source and sampling domains. (a)-(b). Using synthetic data. (c)-(d). Using real data. correction with IS, consistently outperforms wERM+ETIC model on 4 datasets. On Office-31 and Image-CLEF, IS￾based model improves the p…
Figure 7
Figure 7. Figure 7: Two-dimensional visualizations for the source space, sampling space, and their decision boundaries on Office-Home dataset, where different colors [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: t-SNE [61] visualizations of the representations learned by different PDA methods on the Office-Home task Ar→Rw. In the diagrams, “◦” means source or sampling domain, and “+” means target domain; different colors represent different classes, and gray represents outlier…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references · 58 canonical work pages

  1. [1]

    Optimal transport for domain adaptation,

    N. Courty, R. Flamary, D. Tuia et al. , “Optimal transport for domain adaptation,” IEEE Transactions on Pattern Analysis and Machine Intel- ligence, vol. 39, no. 9, pp. 1853–1865, 2017

    work page 2017

  2. [2]

    Domain adaptation via transfer component analysis,

    S. J. Pan, I. W. Tsang, J. T. Kwok et al., “Domain adaptation via transfer component analysis,” IEEE Transactions on Neural Networks , vol. 22, no. 2, pp. 199–210, 2010

    work page 2010

  3. [3]

    Source data-absent unsupervised do- main adaptation through hypothesis transfer and labeling transfer,

    J. Liang, D. Hu, Y . Wang et al., “Source data-absent unsupervised do- main adaptation through hypothesis transfer and labeling transfer,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 44, no. 11, pp. 8602–8617, 2021

    work page 2021

  4. [4]

    A theory of learning from different domains,

    S. Ben David, J. Blitzer, K. Crammer et al., “A theory of learning from different domains,” Machine Learning, vol. 79, pp. 151–175, 2010

    work page 2010

  5. [5]

    Impossibility theorems for domain adaptation,

    S. Ben David, T. Lu, T. Luu et al., “Impossibility theorems for domain adaptation,” in Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics . JMLR Workshop and Confer- ence Proceedings, 2010, pp. 129–136

    work page 2010

  6. [6]

    Integrating structured biological data by kernel maximum mean discrepancy,

    K. M. Borgwardt, A. Gretton, M. J. Rasch et al., “Integrating structured biological data by kernel maximum mean discrepancy,” Bioinformatics, vol. 22, no. 14, pp. e49–e57, 2006

    work page 2006

  7. [7]

    Learning transferable features with deep adaptation networks,

    M. Long, Y . Cao, J. Wang et al. , “Learning transferable features with deep adaptation networks,” in International Conference on Machine Learning. PMLR, 2015, pp. 97–105

    work page 2015

  8. [8]

    Deep subdomain adaptation network for image classification,

    Y . Zhu, F. Zhuang, J. Wang et al., “Deep subdomain adaptation network for image classification,” IEEE Transactions on Neural Networks and Learning Systems, vol. 32, no. 4, pp. 1713–1722, 2020

    work page 2020

Show all 62 references
  1. [9]

    Return of frustratingly easy domain adaptation,

    B. Sun, J. Feng, and K. Saenko, “Return of frustratingly easy domain adaptation,” in Proceedings of the AAAI Conference on Artificial Intel- ligence, vol. 30, no. 1, 2016, pp. 2058–2065

    2016

  2. [10]

    Geodesic flow kernel for unsupervised domain adaptation,

    B. Gong, Y . Shi, F. Sha et al., “Geodesic flow kernel for unsupervised domain adaptation,” in 2012 IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2012, pp. 2066–2073

    2012

  3. [11]

    Heterogeneous domain adaptation via covariance structured feature translators,

    C. X. Ren, J. Feng, D. Q. Dai et al., “Heterogeneous domain adaptation via covariance structured feature translators,” IEEE Transactions on Cybernetics, vol. 51, no. 4, pp. 2166–2177, 2019

    2019

  4. [12]

    Unsupervised domain adaptation via discriminative manifold propagation,

    Y . W. Luo, C. X. Ren, D. Q. Daiet al., “Unsupervised domain adaptation via discriminative manifold propagation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 3, pp. 1653–1669, 2020

    2020

  5. [13]

    Domain-adversarial training of neural networks,

    Y . Ganin, E. Ustinova, H. Ajakan et al., “Domain-adversarial training of neural networks,” The Journal of Machine Learning Research , vol. 17, no. 1, pp. 2096–2030, 2016

    2016

  6. [14]

    Adversarial discriminative do- main adaptation,

    E. Tzeng, J. Hoffman, K. Saenko et al., “Adversarial discriminative do- main adaptation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2017, pp. 7167–7176

    2017

  7. [15]

    Discriminative adversarial domain adaptation,

    H. Tang and K. Jia, “Discriminative adversarial domain adaptation,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 34, no. 04, 2020, pp. 5940–5947

    2020

  8. [16]

    Domain adaptive person re- identification via camera style generation and label propagation,

    C. X. Ren, B. Liang, P. Ge et al. , “Domain adaptive person re- identification via camera style generation and label propagation,” IEEE Transactions on Information Forensics and Security , vol. 15, pp. 1290– 1302, 2019

    2019

  9. [17]

    Joint distribution optimal transportation for domain adaptation,

    N. Courty, R. Flamary, A. Habrard et al. , “Joint distribution optimal transportation for domain adaptation,” Advances in Neural Information Processing Systems, vol. 30, pp. 3733–3742, 2017

    2017

  10. [18]

    Deepjdot: Deep joint distribution optimal transport for unsupervised domain adapta- tion,

    B. B. Damodaran, B. Kellenberger, R. Flamary et al., “Deepjdot: Deep joint distribution optimal transport for unsupervised domain adapta- tion,” in Proceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 447–463

    2018

  11. [19]

    Optimal transport in reproducing kernel hilbert spaces: Theory and applications,

    Z. Zhang, M. Wang, and A. Nehorai, “Optimal transport in reproducing kernel hilbert spaces: Theory and applications,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 42, no. 7, pp. 1741– 1754, 2019

    2019

  12. [20]

    Generalized conditional domain adaptation: A causal perspective with low-rank translators,

    C. X. Ren, X. L. Xu, and H. Yan, “Generalized conditional domain adaptation: A causal perspective with low-rank translators,” IEEE Trans- actions on Cybernetics , vol. 50, no. 2, pp. 821–834, 2018

    2018

  13. [21]

    Mapping conditional distributions for domain adaptation under generalized target shift,

    M. Kirchmeyer, A. Rakotomamonjy, E. de Bezenac et al. , “Mapping conditional distributions for domain adaptation under generalized target shift,” arXiv preprint arXiv:2110.15057 , 2021

    2021 arXiv

  14. [22]

    Domain adaptation with conditional distribution matching and generalized label shift,

    R. Tachet des Combes, H. Zhao, Y . X. Wang et al., “Domain adaptation with conditional distribution matching and generalized label shift,” Ad- vances in Neural Information Processing Systems , vol. 33, pp. 19 276– 19 289, 2020

    2020

  15. [23]

    Domain adaptation under target and conditional shift,

    K. Zhang, B. Sch ¨olkopf, K. Muandet et al., “Domain adaptation under target and conditional shift,” in International Conference on Machine Learning. PMLR, 2013, pp. 819–827

    2013

  16. [24]

    Generalized label shift correction via minimum uncertainty principle: Theory and algorithm,

    Y . W. Luo and C. X. Ren, “Generalized label shift correction via minimum uncertainty principle: Theory and algorithm,” arXiv preprint arXiv:2202.13043, 2022

    2022 arXiv

  17. [25]

    Learning to transfer examples for partial domain adaptation,

    Z. Cao, K. You, M. Long et al. , “Learning to transfer examples for partial domain adaptation,” in Proceedings of the IEEE/CVF Conference On Computer Vision and Pattern Recognition , 2019, pp. 2985–2994

    2019

  18. [26]

    Deep residual correction network for partial domain adaptation,

    S. Li, C. H. Liu, Q. Lin et al. , “Deep residual correction network for partial domain adaptation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 7, pp. 2329–2344, 2020

    2020

  19. [27]

    Mind the class weight bias: Weighted maximum mean discrepancy for unsupervised domain adaptation,

    H. Yan, Y . Ding, P. Li et al. , “Mind the class weight bias: Weighted maximum mean discrepancy for unsupervised domain adaptation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 2272–2281

    2017

  20. [28]

    Partial adversarial domain adaptation,

    Z. Cao, L. Ma, M. Long et al., “Partial adversarial domain adaptation,” in Proceedings of the European Conference on Computer Vision (ECCV) , 2018, pp. 135–150

    2018

  21. [29]

    Associative partial domain adaptation,

    Y . Kim, S. Hong, S. Yang et al., “Associative partial domain adaptation,” arXiv preprint arXiv:2008.03111 , 2020

    2008 arXiv

  22. [30]

    Unsupervised domain adaptation via deep conditional adaptation network,

    P. Ge, C. X. Ren, X. L. Xu et al., “Unsupervised domain adaptation via deep conditional adaptation network,” Pattern Recognition, vol. 134, p. 109088, 2023

    2023

  23. [31]

    Conditional adversarial domain adap- tation,

    M. Long, Z. Cao, J. Wang et al., “Conditional adversarial domain adap- tation,” Advances in Neural Information Processing Systems , vol. 31, pp. 1647–1657, 2018

    2018

  24. [32]

    Adversarial domain adaptation with domain mixup,

    M. Xu, J. Zhang, B. Ni et al. , “Adversarial domain adaptation with domain mixup,” in Proceedings of the AAAI conference on Artificial Intelligence, vol. 34, no. 04, 2020, pp. 6502–6509

    2020

  25. [33]

    Enhanced transport distance for unsupervised domain adaptation,

    M. Li, Y . M. Zhai, Y . W. Luo et al. , “Enhanced transport distance for unsupervised domain adaptation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 13 936–13 944

    2020

  26. [34]

    Conditional bures metric for domain adaptation,

    Y . W. Luo and C. X. Ren, “Conditional bures metric for domain adaptation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 13 989–13 998

    2021

  27. [35]

    Buresnet: Conditional bures metric for transferable representation learning,

    C. X. Ren, Y . W. Luo, and D. Q. Dai, “Buresnet: Conditional bures metric for transferable representation learning,” IEEE Transactions on JOURNAL OF LATEX CLASS FILES, VOL. 18, NO. 9, SEPTEMBER 2020 14 Pattern Analysis and Machine Intelligence , vol. 45, no. 4, pp. 4198– 4213, 2022

    2020

  28. [36]

    Larger norm more transferable: An adaptive feature norm approach for unsupervised domain adaptation,

    R. Xu, G. Li, J. Yang et al. , “Larger norm more transferable: An adaptive feature norm approach for unsupervised domain adaptation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 1426–1435

    2019

  29. [37]

    Contrastive vicinal space for unsupervised domain adaptation,

    J. Na, D. Han, H. J. Chang et al. , “Contrastive vicinal space for unsupervised domain adaptation,” in European Conference on Computer Vision. Springer, 2022, pp. 92–110

    2022

  30. [38]

    Adversarial reweighting for partial do- main adaptation,

    X. Gu, X. Yu, J. Sun et al. , “Adversarial reweighting for partial do- main adaptation,” Advances in Neural Information Processing Systems , vol. 34, pp. 14 860–14 872, 2021

    2021

  31. [39]

    Improving mini-batch optimal transport via partial transportation,

    K. Nguyen, D. Nguyen, T. Pham et al., “Improving mini-batch optimal transport via partial transportation,” in International Conference on Machine Learning. PMLR, 2022, pp. 16 656–16 690

    2022

  32. [40]

    MOT: Masked optimal transport for partial domain adaptation,

    Y . W. Luo and C. X. Ren, “MOT: Masked optimal transport for partial domain adaptation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 3531–3540

    2023

  33. [41]

    Select, label, and mix: Learn- ing discriminative invariant feature representations for partial domain adaptation,

    A. Sahoo, R. Panda, R. Feris et al. , “Select, label, and mix: Learn- ing discriminative invariant feature representations for partial domain adaptation,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2023, pp. 4210–4219

    2023

  34. [42]

    Reinforced adaptation network for partial domain adaptation,

    K. Wu, M. Wu, Z. Chen, et al. , “Reinforced adaptation network for partial domain adaptation,” IEEE Transactions on Circuits and Systems for Video Technology, 2022

    2022

  35. [43]

    From big to small: Adaptive learning to partial-set domains,

    Z. Cao, K. You, Z. Zhang, et al., “From big to small: Adaptive learning to partial-set domains,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 2, pp. 1766–1780, 2022

    2022

  36. [44]

    Partial domain adaptation without domain align- ment,

    W. Li and S. Chen, “Partial domain adaptation without domain align- ment,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022

    2022

  37. [45]

    Small is beautiful: Compressing deep neural networks for partial domain adaptation,

    Y . Ma, X. Yao, R. Chen et al. , “Small is beautiful: Compressing deep neural networks for partial domain adaptation,” IEEE Transactions on Neural Networks and Learning Systems , vol. 35, no. 3, pp. 3575–3585, 2024

    2024

  38. [46]

    Adversarial partial domain adaptation by cycle inconsistency,

    K. Y . Lin, J. Zhou, Y . Qiu,et al., “Adversarial partial domain adaptation by cycle inconsistency,” in European Conference on Computer Vision . Springer, 2022, pp. 530–548

    2022

  39. [47]

    Importance sampling: a review,

    S. T. Tokdar and R. E. Kass, “Importance sampling: a review,” Wiley Interdisciplinary Reviews: Computational Statistics , vol. 2, no. 1, pp. 54–60, 2010

    2010

  40. [48]

    On mixup regularization,

    L. Carratino, M. Ciss ´e, R. Jenatton et al. , “On mixup regularization,” arXiv preprint arXiv:2006.06049 , 2020

    2006 arXiv

  41. [49]

    Entropy regularized optimal trans- port independence criterion,

    L. Liu, S. Pal, and Z. Harchaoui, “Entropy regularized optimal trans- port independence criterion,” in International Conference on Artificial Intelligence and Statistics . PMLR, 2022, pp. 11 247–11 279

    2022

  42. [50]

    Kernel methods for measuring independence,

    A. Gretton, R. Herbrich, A. Smola et al., “Kernel methods for measuring independence,” Journal of Machine Learning Research , vol. 6, no. 70, pp. 2075–2129, 2005

    2005

  43. [51]

    A kernel statistical test of independence,

    A. Gretton, K. Fukumizu, C. Teo et al. , “A kernel statistical test of independence,” Advances in Neural Information Processing Systems , vol. 20, 2007

    2007

  44. [52]

    Interpolating between optimal transport and mmd using sinkhorn divergences,

    J. Feydy, T. S ´ejourn´e, F. X. Vialardet al., “Interpolating between optimal transport and mmd using sinkhorn divergences,” in The 22nd Interna- tional Conference on Artificial Intelligence and Statistics. PMLR, 2019, pp. 2681–2690

    2019

  45. [53]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren et al. , “Deep residual learning for image recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 770–778

    2016

  46. [54]

    Detecting and correcting for label shift with black box predictors,

    Z. Lipton, Y . X. Wang, and A. Smola, “Detecting and correcting for label shift with black box predictors,” in International Conference on Machine Learning. PMLR, 2018, pp. 3122–3130

    2018

  47. [55]

    Imagenet: A large-scale hierar- chical image database,

    J. Deng, W. Dong, R. Socher, et al. , “Imagenet: A large-scale hierar- chical image database,” in IEEE Conference on Computer Vision and Pattern Recognition. Ieee, 2009, pp. 248–255

    2009

  48. [56]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  49. [57]

    Imageclef 2014: Overview and analysis of the results,

    B. Caputo, H. M ¨uller, J. Martinez Gomez et al. , “Imageclef 2014: Overview and analysis of the results,” in Information Access Evalua- tion. Multilinguality, Multimodality, and Interaction: 5th International Conference of the CLEF Initiative, CLEF 2014, Sheffield, UK, Septem...

    2014

  50. [58]

    Adapting visual category models to new domains,

    K. Saenko, B. Kulis, M. Fritz et al., “Adapting visual category models to new domains,” in Computer Vision–ECCV 2010: 11th European Conference on Computer Vision, Heraklion, Crete, Greece, September 5-11, 2010, Proceedings, Part IV 11 . Springer, 2010, pp. 213–226

    2010

  51. [59]

    Deep hashing network for unsupervised domain adaptation,

    H. Venkateswara, J. Eusebio, S. Chakraborty et al. , “Deep hashing network for unsupervised domain adaptation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2017, pp. 5018–5027

    2017

  52. [60]

    Visda: The visual domain adaptation challenge,

    X. Peng, B. Usman, N. Kaushik et al. , “Visda: The visual domain adaptation challenge,” arXiv preprint arXiv:1710.06924 , 2017

    2017 arXiv

  53. [61]

    Visualizing data using t-sne

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.” Journal of Machine Learning Research , vol. 9, no. 11, 2008

    2008

  54. [62]

    mixup: Beyond empirical risk minimization,

    H. Zhang, “mixup: Beyond empirical risk minimization,” in Interna- tional Conference on Learning Representations , 2018, pp. 1–13

    2018

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.