Pith. sign in

REVIEW 2 major objections 4 minor 58 references

Theoretical Performance Guarantees for Partial Domain Adaptation via Partial Optimal Transport

T0 review · 2 major / 4 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 Partial optimal transport gives partial domain adaptation a theorem-backed way to weight source samples and align features.

desk verdict Real theoretical contribution on partial Wasserstein alignment, but the paper overclaims the link to its cross-entropy objective. read the letter →

arxiv 2506.02712 v1 pith:CKMWDMXO submitted 2025-06-03 cs.LG stat.ML

Jayadev Naram , Fredrik Hellström , Ziming Wang , Rebecka Jörnsten , Giuseppe Durisi This is my paper
classification cs.LGstat.ML
keywords partialdomainadaptationoptimaltransportWassersteindistancegeneralizationboundsnegativetransfersourcereweightingPAC-Bayesunsupervised
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 sets out to show that partial domain adaptation—learning from a source domain whose label set contains the target label set—has a first-principles justification in partial optimal transport. It proves upper bounds on the target loss in which the alignment term is a partial Wasserstein distance and the source losses are weighted by the row sums of the optimal transport coupling, so the weights are derived quantities rather than heuristics. A second, joint feature-label version of the bound, followed by a PAC-Bayes argument that adds a KL-divergence term, lifts the result to high-probability generalization bounds. If the argument is right, two common practices, aligning feature distributions and reweighting source samples to reduce negative transfer, become components of a provable bound with an explicit recipe for the weights. The paper also presents WARMPOT as a practical minimizer of these terms, noting that exact minimization is prohibitively expensive and that its implementation relies on entropic mini-batch partial transport.

What carries the argument

The engine is the partial Wasserstein distance $\mathrm{PW}_\alpha(1/\beta P_s^f, Q_t^f)$: an optimal transport problem that moves only $\alpha$ units of mass under the marginal constraints $\Pi 1_{n_t} \le P_s^f/\beta$ and $\Pi^T 1_{n_s} \le Q_t^f$. Its optimal coupling $\Pi^\star$ simultaneously supplies the alignment cost and, through its row and column sums, the source and target weights $p_i, q_j$ that reweight the empirical losses. Two auxiliary devices carry the proof: a change-of-measure inequality that converts the unevenly weighted target loss back to the uniform empirical target loss at the price of a total-variation term, and a difficulty term $L_f$ (or $\hat L_f$) that measures the best achievable loss on the combined labeled data. A PAC-Bayes lemma then lifts the fixed-sample inequality to a high-probability bound on the population target loss, adding a KL-divergence term and a variance term of order $1/n_t$.

What would settle it

On synthetic data with known target labels and a metric label loss, compute the empirical target loss and the right-hand side of the bound in Theorem 3.2 for many fixed $\gamma$-Lipschitz classifiers and feature maps; any violation of the inequality would refute the theorem. For the algorithmic claim, train the same model with cross-entropy and with a metric label loss; if the metric-loss version does not match or improve on cross-entropy, the experimental success is not explained by the paper's assumptions.

Watch

Extended reading notes

Core claim

The paper's central claim is that, with a fixed feature extractor $f$ and any classifier $g$ that is $\gamma$-Lipschitz with respect to a metric label loss $\ell$, the empirical target loss obeys $$L_{\tilde z}(w) \le \sum_{i=1}^{n_s} \frac{p_i}{\$\alpha$} \ell(w(x_i), y_i) + \frac{2}{\$\alpha$} \mathrm{PW}_\$\alpha$\left(\frac{1}{\$\beta$} P_s^f, Q_t^f\right) + \frac{1}{2}\sum_{j=1}^{n_t} \left|\frac{1}{n_t} - \frac{q_j}{\$\alpha$}\right| + 2L_f,$$ where $p_i$ and $q_j$ are the row and column sums of the optimal coupling matrix $\Pi^\star$ of the partial Wasserstein problem and $L_f$ is a non-computable term measuring problem difficulty. A second bound replaces the feature-only distributions by joint feature-label distributions and adds a label cost to the transport metric. The paper proves both bounds and converts them into PAC-Bayes generalization bounds, emphasizing that the partial Wasserstein term is computed directly on empirical feature distributions, so the bound explicitly depends on the learned feature map and motivates feature alignment.

Load-bearing premise

The load-bearing premise is that the label loss is a true distance between labels (symmetric and satisfying the triangle inequality) and, for the joint version, Lipschitz in each argument; the algorithm and experiments instead use cross-entropy, which is neither, and the paper does not close that gap.

Editorial extensions

If this is right

  • The weights in a weighted source loss are no longer free: the optimal partial transport coupling prescribes them, so any PDA method using a weighted source term has a principled default.
  • The partial Wasserstein distance with its two budget parameters is supported as a domain alignment term, with $\beta$ controlling the inflated source mass and $\alpha$ controlling how much target mass participates.
  • Because the bound is evaluated on empirical feature distributions, it explicitly depends on the learned feature map, motivating training the feature extractor and not just the classifier.
  • The coupling automatically suppresses negative transfer: in the paper's P→A experiment, outlier source classes receive only about 6% of the source weight mass.
  • When the theoretically derived weights are used inside the ARPM algorithm, the paper reports the highest average Office-Home accuracy among the compared methods (82.0%), suggesting the weight recipe transfers beyond the paper's own objective.

Reading between the lines

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

  • The theorems assume a metric label loss, while the experiments train with cross-entropy, which is neither a metric nor Lipschitz in the label argument; whether the bounds extend to such losses is unsettled, and training with a metric label cost would test how much theory covers the deployed objective.
  • The per-sample weights double as outlier scores: a source point with near-zero $p_i$ is one the transport plan is willing to discard, so the same coupling could drive sample selection, open-set adaptation, or robust training.
  • Because the bound depends explicitly on the feature map $f$, it could serve as a model-selection or training criterion for feature extractors, not only as a justification of alignment; the paper does not develop this use.
  • Appendix H shows that with $\alpha=1$ the WARMPOT weights solve a Wasserstein-1 problem with the box constraint $\hat p_i \le 1/(\beta n_s)$, in contrast to ARPM's variance constraint, suggesting box-constrained transport as a general template for reweighting schemes in other domain adaptation settings.
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

2 major / 4 minor

Summary. The paper derives two families of generalization bounds for partial domain adaptation (PDA) based on partial optimal transport. For a fixed hypothesis w = g∘f, Theorem 3.2 bounds the empirical target loss by a weighted empirical source loss, a partial Wasserstein distance between empirical source and target feature distributions, a total-variation term, and a non-computable difficulty term. Theorem 3.3 gives an analogous bound using joint feature-label distributions, with the target labels replaced by predictions under w. In both bounds, the source weights pi and target weights qj are the row and column sums of the optimal partial-transport coupling matrix. Lemma 3.4 converts these fixed-hypothesis bounds into PAC-Bayes generalization bounds (Corollaries 3.5 and 3.6). Based on these results, the authors propose the WARMPOT algorithm that minimizes a weighted cross-entropy source loss plus a partial Wasserstein alignment term computed with cross-entropy in the transport cost, and they report experiments on Office-Home and ImageNet-to-Caltech showing that WARMPOT is competitive with recent methods and that replacing the ARPM weights with the WARMPOT weights improves ARPM's average accuracy.

Significance. If the bounds are valid, they provide a genuinely new theoretical foundation for two widely used ingredients in PDA algorithms: the partial Wasserstein distance as a domain-alignment term and non-uniform weighting of source samples. A notable strength is that the Wasserstein term in the bounds is expressed directly through empirical measures, avoiding the additional concentration-of-measure step present in earlier OT-based domain-adaptation bounds. The weights arise constructively from the optimal coupling rather than being imposed heuristically, and the authors ship an open-source implementation. However, the theoretical results are stated under a metric loss bounded in [0,1], while the deployed objective uses cross-entropy, which is neither a metric nor bounded. This gap means that the theorems, as stated, do not cover the actual quantity minimized by WARMPOT. The paper also lacks a clear validation protocol for its tuned hyperparameters. For these reasons the significance of the results is real but conditional on closing the theory-practice gap.

major comments (2)
  1. [Section 5.2; Appendix E] There is a fundamental mismatch between the loss covered by the theory and the loss used in the algorithm. Theorems 3.2 and 3.3 assume that the loss function ℓ is a metric on Y and bounded in [0,1]; Theorem 3.3 further requires ζ-Lipschitzness in each argument. Section 5.2 states that WARMPOT sets ℓ to the cross-entropy loss in the objective (19), both in the weighted source term and inside the partial-transport cost. Cross-entropy is not a metric (it is not symmetric and does not satisfy the triangle inequality; it is also unbounded above on the probability simplex). Consequently, the proofs in Appendix B and C, which rely on the triangle and reverse-triangle inequalities (Lemma A.4, Eq. (26); Eq. (59)), and the Hoeffding step in Lemma 3.4 (which requires ℓ∈[0,1]), do not apply to the objective actually minimized. The paper's limitation paragraph in Section 6 acknowledges only the computational difficulty of exact bound minimization, not this loss mismatch. The authors should either extend the theorems to a class of losses that includes cross-entropy (with boundedness and smoothness assumptions appropriate for softmax outputs) or modify WARMPOT to employ a loss satisfying the metric and boundedness assumptions, and then discuss the implications for the theoretical motivation of the algorithm.
  2. [Appendix E] The hyperparameters (αmax, β, η1, η2, ε) are said to be obtained 'through a parameter search', but the manuscript does not specify whether this search was performed on a separate validation split or directly on the test tasks. Without a clear validation protocol, the reader cannot assess whether the reported competitive results are affected by test-set feedback. Please state explicitly the split used for hyperparameter selection and, if any tuning was performed on the test tasks, provide an evaluation in which hyperparameters are selected without access to test labels, or report the sensitivity of the conclusions to this choice.
minor comments (4)
  1. [Table 1] The comparison of weighting schemes is potentially confounded by different weight-update frequencies: WARMPOT updates weights every mini-batch, while BA3US and ARPM update them only every 500 iterations on the full dataset. A control experiment with matched update schedules would strengthen the claim that the weighting scheme itself, rather than the update frequency, drives the observed improvement.
  2. [Table 2] The claim that ARPM+our-weights achieves state-of-the-art performance rests on a 0.3 percentage-point average improvement over ARPM (82.0 vs 81.7 with standard deviations of 0.4 and 0.7). A paired significance test or confidence intervals over the common seeds would make the superiority claim more robust.
  3. [Theorem 3.3] The definition of L̂f in (13) with Ξ in (63) is difficult to parse because it mixes minimizations over g′ with the coupling-dependent weights. A short intuitive explanation of why this term is non-computable and what it represents in terms of the difficulty of the PDA problem would improve readability.
  4. [Various] There are several typographical errors, e.g., 'Wassserstein' in Section 1 ('Our proof techniques ... Wassserstein distance') and the inconsistent spelling of 'WARMPOT' in Table captions. A careful proofreading pass is recommended.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the bounds are proven inequalities for fixed hypotheses, and the optimal-transport weights are derived in the proof rather than fitted to the target loss.

full rationale

The paper's central derivation is self-contained. Theorem 3.2 starts from Lemma A.4, a Lipschitz inequality for the metric loss, and multiplies both sides by an optimal coupling Π* of PW_α(1/β P_s^f, Q_t^f). Rearranging the resulting sum yields the bound (5), with pi and qj defined as the row and column sums of Π*. This is a proof technique, not a circular definition: the target loss L_˜z(w) on the left is not used to define the weights, and the bound holds for every fixed w in the stated class. Theorem 3.3 and the PAC-Bayes Corollaries 3.5–3.6 follow the same pattern: the joint-distribution PW term depends on w only through predicted target labels, and Lemma 3.4 is proved in Appendix D from Hoeffding and Donsker-Varadhan. There is no fitted parameter renamed as a prediction, and no load-bearing self-citation: references to Hellström et al. (2025) and Wang et al. (2024) are contextual or for comparison, and the key proof relies on Alquier (2024) and standard results. The main weakness is a scope gap—Theorems 3.2/3.3 assume a bounded metric loss while WARMPOT uses cross-entropy—but that is a correctness/validity concern, not circularity, because the theorems do not assume the conclusion they purport to establish. Hence no circular step can be exhibited under the required standard.

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

The bounds rest on standard metric and Lipschitz assumptions and the PAC-Bayes machinery; the algorithm additionally depends on five hand-tuned hyperparameters. No invented entities are introduced.

free parameters (5)
  • alpha (α) = 0.8 (Office-Home), 0.08 (ImageNet->Caltech), linearly annealed from 0.01
    Chosen by parameter search in Appendix E/F; controls the fraction of target mass transported and appears in the bound's normalizing factors.
  • beta (β) = 0.35 (Office-Home), 0.72 (ImageNet->Caltech)
    Scales the source distribution by 1/beta; hand-tuned and not predicted by the theory.
  • eta1 (η1) = 0.125 (Office-Home), 0.92 (ImageNet->Caltech)
    Weight on inter-feature distance in the partial Wasserstein cost; chosen by parameter search.
  • eta2 (η2) = 1.75 (Office-Home), 5.47 (ImageNet->Caltech)
    Weight on inter-label distance in the joint cost; chosen by parameter search.
  • epsilon (ε) = 7.0 (Office-Home), 5.59 (ImageNet->Caltech)
    Entropic regularization constant in the OT solver; chosen to avoid numerical instabilities.
assumptions (5)
  • domain assumption The loss ℓ is a metric on Y (Thm 3.2) and ζ-Lipschitz in each argument (Thm 3.3).
    Needed for Lemma A.4 and Theorem 3.3 proof; violated by the cross-entropy loss used in experiments.
  • domain assumption Classifiers g in W are γ-Lipschitz with respect to the relevant metric.
    Used in Lemma A.4 and in the cost of the partial Wasserstein term; no Lipschitz constant is estimated for the networks used in experiments.
  • domain assumption Loss is bounded in [0,1] and the feature extractor f is fixed and measurable.
    Boundedness is used for Hoeffding in Lemma 3.4 and for the TV change of measure; measurability is assumed throughout.
  • standard math PAC-Bayes bound of Lemma 3.4 and Donsker-Varadhan variational formula hold.
    Used to convert pointwise empirical-target-loss bounds into high-probability population-loss bounds.
  • standard math Optimal couplings for the partial OT problem exist for finite discrete measures.
    Used to define weights pi and qj; holds under compactness and continuity for the discrete problem.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Theoretical Performance Guarantees for Partial Domain Adaptation via Partial Optimal Transport." pith.science (2026). https://pith.science/paper/CKMWDMXO

@misc{pith2026250602712,
  author       = {Pith},
  title        = {Pith review of: Theoretical Performance Guarantees for Partial Domain Adaptation via Partial Optimal Transport},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CKMWDMXO}},
  note         = {Machine review of arXiv:2506.02712}
}
read the original abstract

In many scenarios of practical interest, labeled data from a target distribution are scarce while labeled data from a related source distribution are abundant. One particular setting of interest arises when the target label space is a subset of the source label space, leading to the framework of partial domain adaptation (PDA). Typical approaches to PDA involve minimizing a domain alignment term and a weighted empirical loss on the source data, with the aim of transferring knowledge between domains. However, a theoretical basis for this procedure is lacking, and in particular, most existing weighting schemes are heuristic. In this work, we derive generalization bounds for the PDA problem based on partial optimal transport. These bounds corroborate the use of the partial Wasserstein distance as a domain alignment term, and lead to theoretically motivated explicit expressions for the empirical source loss weights. Inspired by these bounds, we devise a practical algorithm for PDA, termed WARMPOT. Through extensive numerical experiments, we show that WARMPOT is competitive with recent approaches, and that our proposed weights improve on existing schemes.

Figures

Figures reproduced from arXiv: 2506.02712 by the authors.

Figure 1
Figure 1. The distribution of WARMPOT weights for the task P→A. Most of the weights of the outlier classes are close to zero, suggesting that most of the outliers are successfully omitted when training the classifier. specifically highlighted here. References Alquier, P. User-friendly introduction to PAC-Bayes bounds. Found. Trends Mach. Learn., 17(2):174–303, Jan. 2024. Arjovsky, M., Chintala, S., and Bottou, L. Wasserstein … view at source ↗
Figure 2
Figure 2. The effect of changing alignment parameters αmax and β on test accuracy of ImageNet → Caltech. results indicate that the specific choice of these parameters has a minor impact over a range of reasonable values. H. Relation between WARMPOT and ARPM Weights The weights pˆ = (ˆp1, . . . , pˆns ) used in the ARPM algorithm of Gu et al. (2024) are defined as the solution of the following Wasserstein-1 type problem betwee… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

58 extracted references · 58 canonical work pages

  1. [1]

    User-friendly introduction to PAC-Bayes bounds

    Alquier, P. User-friendly introduction to PAC-Bayes bounds. Found. Trends Mach. Learn., 17 0 (2): 0 174--303, Jan. 2024

    work page 2024

  2. [2]

    Wasserstein generative adversarial networks

    Arjovsky, M., Chintala, S., and Bottou, L. Wasserstein generative adversarial networks. In Proc. Int. Conf. Mach. Learning (ICML), Sydney, Australia, Aug. 2017

    work page 2017

  3. [3]

    Analysis of representations for domain adaptation

    Ben-David, S., Blitzer, J., Crammer, K., and Pereira, F. Analysis of representations for domain adaptation. In Proc. Conf. Neural Inf. Process. Syst. (NeurIPS), Vancouver, Canada, Dec. 2006

    work page 2006

  4. [4]

    Ben-David, S., Blitzer, J., Crammer, K., Kulesza, A., Pereira, F., and Vaughan, J. W. A theory of learning from different domains. Mach. Learn., 79 0 (1): 0 151--175, Oct. 2010

    work page 2010

  5. [5]

    and Vandenberghe, L

    Boyd, S. and Vandenberghe, L. Convex Optimization. Cambridge Univ. Press, Cambridge, UK, 2004

    work page 2004

  6. [6]

    Caffarelli, L. A. and McCann, R. J. Free boundaries in optimal transport and Monge-Ampere obstacle problems. Ann. Math., 171 0 (2): 0 673--730, Mar. 2010

    work page 2010

  7. [7]

    Partial adversarial domain adaptation

    Cao, Z., Ma, L., Long, M., and Wang, J. Partial adversarial domain adaptation. In Proc. Eur. Conf. Comput. Vis. (ECCV), Munich, Germany, Sep. 2018

    work page 2018

  8. [8]

    Learning to transfer examples for partial domain adaptation

    Cao, Z., You, K., Long, M., Wang, J., and Yang, Q. Learning to transfer examples for partial domain adaptation. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Long Beach, CA, USA, June 2019

    work page 2019

Show all 58 references
  1. [9]

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

    Cao, Z., You, K., Zhang, Z., Wang, J., and Long, M. From big to small: Adaptive learning to partial-set domains. IEEE Trans. Pattern Anal. Mach. Intell. (PAMI), 45 0 (2): 0 1766--1780, Mar. 2022

    2022

  2. [10]

    PAC - Bayesian Supervised Classification: the Thermodynamics of Statistical Learning

    Catoni, O. PAC - Bayesian Supervised Classification: the Thermodynamics of Statistical Learning . IMS Lecture Notes Monogr. Ser., 56, Beachwood, OH, USA , 2007

    2007

  3. [11]

    Unified optimal transport framework for universal domain adaptation

    Chang, W., Shi, Y., Tuan, H., and Wang, J. Unified optimal transport framework for universal domain adaptation. In Proc. Conf. Neural Inf. Process. Syst. (NeurIPS), New Orleans, Louisiana, USA, Nov. 2022

    2022

  4. [12]

    Domain adaptation with regularized optimal transport

    Courty, N., Flamary, R., and Tuia, D. Domain adaptation with regularized optimal transport. In Proc. Mach. Learn. Knowl. Discov. Databases (ECML PKDD), Nancy, France, Sep. 2014

    2014

  5. [13]

    Joint distribution optimal transportation for domain adaptation

    Courty, N., Flamary, R., Habrard, A., and Rakotomamonjy, A. Joint distribution optimal transportation for domain adaptation. In Proc. Conf. Neural Inf. Process. Syst. (NeurIPS), Long Beach, CA, USA, Dec. 2017 a

    2017

  6. [14]

    Optimal transport for domain adaptation

    Courty, N., Flamary, R., Tuia, D., and Rakotomamonjy, A. Optimal transport for domain adaptation. IEEE Trans. Pattern Anal. Mach. Intell. (PAMI), 39 0 (9): 0 1853--1865, Oct. 2017 b

    2017

  7. [15]

    B., Kellenberger, B., Flamary, R., Tuia, D., and Courty, N

    Damodaran, B. B., Kellenberger, B., Flamary, R., Tuia, D., and Courty, N. DeepJDOT : Deep joint distribution optimal transport for unsupervised domain adaptation. In Proc. Eur. Conf. Comput. Vis. (ECCV), Munich, Germany, Sep. 2018

    2018

  8. [16]

    Farahani, A., Voghoei, S., Rasheed, K., and Arabnia, H. R. A brief review of domain adaptation. In Adv. Data Sci. Inf. Eng., Springer, Cham, Switzerland , Oct. 2021

    2021

  9. [17]

    Unbalanced minibatch optimal transport; applications to domain adaptation

    Fatras, K., S \'e journ \'e , T., Flamary, R., and Courty, N. Unbalanced minibatch optimal transport; applications to domain adaptation. In Proc. Int. Conf. Mach. Learning (ICML), Virtual Conference, July 2021

    2021

  10. [18]

    The optimal partial transport problem

    Figalli, A. The optimal partial transport problem. Arch. Ration. Mech. Anal., 195 0 (2): 0 533--560, Jan. 2010

    2010

  11. [19]

    Z., Boisbunon, A., Chambon, S., Chapel, L., Corenflos, A., Fatras, K., Fournier, N., Gautheron, L., Gayraud, N

    Flamary, R., Courty, N., Gramfort, A., Alaya, M. Z., Boisbunon, A., Chambon, S., Chapel, L., Corenflos, A., Fatras, K., Fournier, N., Gautheron, L., Gayraud, N. T., Janati, H., Rakotomamonjy, A., Redko, I., Rolet, A., Schutz, A., Seguy, V., Sutherland, D. J., Tavenard, R., Ton...

    2021

  12. [20]

    Domain-adversarial training of neural networks

    Ganin, Y., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., Marchand, M., and Lempitsky, V. Domain-adversarial training of neural networks. J. Mach. Learn. Res. (JMLR), 17 0 (1): 0 2096--2030, Jan. 2016

    2016

  13. [21]

    Caltech 256, Apr

    Griffin, G., Holub, A., and Perona, P. Caltech 256, Apr. 2022

    2022

  14. [22]

    Adversarial reweighting for partial domain adaptation

    Gu, X., Yu, X., Sun, J., Xu, Z., et al. Adversarial reweighting for partial domain adaptation. In Proc. Conf. Neural Inf. Process. Syst. (NeurIPS), Virtual Conference, Dec. 2021

    2021

  15. [23]

    Adversarial reweighting with -power maximization for domain adaptation

    Gu, X., Yu, X., Yang, Y., Sun, J., and Xu, Z. Adversarial reweighting with -power maximization for domain adaptation. Int. J. Comput. Vis. (IJCV), 132 0 (10): 0 4768--4791, May 2024

    2024

  16. [24]

    Selective partial domain adaptation

    Guo, P., Zhu, J., and Zhang, Y. Selective partial domain adaptation. In Proc. Br. Mach. Vis. Conf. (BMVC), London, UK, Nov. 2022

    2022

  17. [25]

    Addressing the overfitting in partial domain adaptation with self-training and contrastive learning

    He, C., Li, X., Xia, Y., Tang, J., Yang, J., and Ye, Z. Addressing the overfitting in partial domain adaptation with self-training and contrastive learning. IEEE Trans. Circuits Syst. Video Technol., 34 0 (3): 0 1532--1545, July 2023

    2023

  18. [26]

    Deep residual learning for image recognition

    He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learning for image recognition. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Las Vegas, USA, June 2016

    2016

  19. [27]

    Generalization bounds: Perspectives from information theory and PAC-Bayes

    Hellström, F., Durisi, G., Guedj, B., and Raginsky, M. Generalization bounds: Perspectives from information theory and PAC-Bayes . Found. Trends Mach. Learn., 18 0 (1): 0 1--223, Jan. 2025

    2025

  20. [28]

    H., Lin, Q., Wen, Q., Su, L., Huang, G., and Ding, Z

    Li, S., Liu, C. H., Lin, Q., Wen, Q., Su, L., Huang, G., and Ding, Z. Deep residual correction network for partial domain adaptation. IEEE Trans. Pattern Anal. Mach. Intell. (PAMI), 43 0 (7): 0 2329--2344, Jan. 2020

    2020

  21. [29]

    and Chen, S

    Li, W. and Chen, S. Partial domain adaptation without domain alignment. IEEE Trans. Pattern Anal. Mach. Intell. (PAMI), 45 0 (7): 0 8787--8797, Dec. 2022

    2022

  22. [30]

    A balanced and uncertainty-aware approach for partial domain adaptation

    Liang, J., Wang, Y., Hu, D., He, R., and Feng, J. A balanced and uncertainty-aware approach for partial domain adaptation. In Proc. Eur. Conf. Comput. Vis. (ECCV), Virtual Conference, Aug. 2020

    2020

  23. [31]

    Source data-absent unsupervised domain adaptation through hypothesis transfer and labeling transfer

    Liang, J., Hu, D., Wang, Y., He, R., and Feng, J. Source data-absent unsupervised domain adaptation through hypothesis transfer and labeling transfer. IEEE Trans. Pattern Anal. Mach. Intell. (PAMI), 44 0 (11): 0 8602--8617, Aug. 2021

    2021

  24. [32]

    Adversarial partial domain adaptation by cycle inconsistency

    Lin, K.-Y., Zhou, J., Qiu, Y., and Zheng, W.-S. Adversarial partial domain adaptation by cycle inconsistency. In Proc. Eur. Conf. Comput. Vis. (ECCV), Tel Aviv, Israel, Oct. 2022

    2022

  25. [33]

    Detecting and correcting for label shift with black box predictors

    Lipton, Z., Wang, Y.-X., and Smola, A. Detecting and correcting for label shift with black box predictors. In Proc. Int. Conf. Mach. Learning (ICML), Stockholm, Sweden, July 2018

    2018

  26. [34]

    Learning transferable features with deep adaptation networks

    Long, M., Cao, Y., Wang, J., and Jordan, M. Learning transferable features with deep adaptation networks. In Proc. Int. Conf. Mach. Learning (ICML), Lille, France, July 2015

    2015

  27. [35]

    Long, M., Cao, Z., Wang, J., and Jordan, M. I. Conditional adversarial domain adaptation. In Proc. Conf. Neural Inf. Process. Syst. (NeurIPS), Montréal, Canada, Dec. 2018

    2018

  28. [36]

    and Ren, C.-X

    Luo, Y.-W. and Ren, C.-X. MOT : Masked optimal transport for partial domain adaptation. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Vancouver, Canada, June 2023

    2023

  29. [37]

    and Ren, C.-X

    Luo, Y.-W. and Ren, C.-X. When invariant representation learning meets label shift: Insufficiency and theoretical insights. IEEE Trans. Pattern Anal. Mach. Intell. (PAMI), 46 0 (12): 0 9407--9422, June 2024

    2024

  30. [38]

    McAllester, D. A. PAC-Bayesian model averaging. In Proc. Conf. Comput. Learn. Theory (COLT), Santa Cruz, CA, USA, July 1999

    1999

  31. [39]

    and Kolter, J

    Nagarajan, V. and Kolter, J. Z. Uniform convergence may be unable to explain generalization in deep learning. In Proc. Conf. Neural Inf. Process. Syst. (NeurIPS), Vancouver, Canada, Dec. 2019

    2019

  32. [40]

    Improving mini-batch optimal transport via partial transportation

    Nguyen, K., Nguyen, D., Pham, T., Ho, N., et al. Improving mini-batch optimal transport via partial transportation. In Proc. Int. Conf. Mach. Learning (ICML), Baltimore, MA , USA, July 2022

    2022

  33. [41]

    and Honorio, J

    Ohnishi, Y. and Honorio, J. Novel change of measure inequalities with applications to PAC-Bayesian bounds and Monte Carlo estimation. In Proc. Int. Conf. Artif. Intell. Stat. (AISTATS), Virtual Conference, Apr. 2021

    2021

  34. [42]

    and Gall, J

    Panareda Busto, P. and Gall, J. Open set domain adaptation. In Proc. IEEE Int. Conf. Comput. Vis. (ICCV), Venice, Italy, Oct. 2017

    2017

  35. [43]

    Theoretical analysis of domain adaptation with optimal transport

    Redko, I., Habrard, A., and Sebban, M. Theoretical analysis of domain adaptation with optimal transport. In Proc. Mach. Learn. Knowl. Discov. Databases (ECML PKDD), Skopje, Macedonia, Sep. 2017

    2017

  36. [44]

    Advances in domain adaptation theory

    Redko, I., Morvant, E., Habrard, A., Sebban, M., and Bennani, Y. Advances in domain adaptation theory. Elsevier, Oxford, UK, 2019

    2019

  37. [45]

    ImageNet large scale visual recognition challenge

    Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. ImageNet large scale visual recognition challenge. Int. J. Comput. Vis. (IJCV), 115 0 (3): 0 211--252, Apr. 2015

    2015

  38. [46]

    Select, label, and mix: Learning discriminative invariant feature representations for partial domain adaptation

    Sahoo, A., Panda, R., Feris, R., Saenko, K., and Das, A. Select, label, and mix: Learning discriminative invariant feature representations for partial domain adaptation. In Proc. IEEE/CVF Winter Conf. Appl. Comput. Vis., Waikoloa, Hawaii, USA, Jan. 2023

    2023

  39. [47]

    Wasserstein distance guided representation learning for domain adaptation

    Shen, J., Qu, Y., Zhang, W., and Yu, Y. Wasserstein distance guided representation learning for domain adaptation. In Proc. AAAI Conf. Artif. Intell. (AAAI), New Orleans, Louisiana, USA, Apr. 2018

    2018

  40. [48]

    Tachet des Combes, R., Zhao, H., Wang, Y.-X., and Gordon, G. J. Domain adaptation with conditional distribution matching and generalized label shift. In Proc. Conf. Neural Inf. Process. Syst. (NeurIPS), Virtual Conference, Dec. 2020

    2020

  41. [49]

    Adversarial discriminative domain adaptation

    Tzeng, E., Hoffman, J., Saenko, K., and Darrell, T. Adversarial discriminative domain adaptation. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Honolulu, Hawaii, USA, July 2017

    2017

  42. [50]

    Deep hashing network for unsupervised domain adaptation

    Venkateswara, H., Eusebio, J., Chakraborty, S., and Panchanathan, S. Deep hashing network for unsupervised domain adaptation. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Honolulu, Hawaii, USA, July 2017

    2017

  43. [51]

    Wainwright, M. J. High-Dimensional Statistics: a Non-Asymptotic Viewpoint. Cambridge Univ. Press, Cambridge, U.K., 2019

    2019

  44. [52]

    Partial Wasserstein adversarial network for non-rigid point set registration

    Wang, Z.-M., Xue, N., Lei, L., and Xia, G.-S. Partial Wasserstein adversarial network for non-rigid point set registration. In Proc. Int. Conf. Learn. Represent. (ICLR), Virtual Conference, Apr. 2022

    2022

  45. [53]

    Partial distribution matching via partial Wasserstein adversarial networks

    Wang, Z.-M., Xue, N., Lei, L., J \"o rnsten, R., and Xia, G.-S. Partial distribution matching via partial Wasserstein adversarial networks. arXiv, Sep. 2024

    2024

  46. [54]

    Reinforced adaptation network for partial domain adaptation

    Wu, K., Wu, M., Chen, Z., Jin, R., Cui, W., Cao, Z., and Li, X. Reinforced adaptation network for partial domain adaptation. IEEE Trans. Circuits Syst. Video Technol., 33 0 (5): 0 2370--2380, Nov. 2023

    2023

  47. [55]

    Implicit semantic response alignment for partial domain adaptation

    Xiao, W., Ding, Z., and Liu, H. Implicit semantic response alignment for partial domain adaptation. In Proc. Conf. Neural Inf. Process. Syst. (NeurIPS), Virtual Conference, Dec. 2021

    2021

  48. [56]

    C., Xue, B., and Zhang, M

    Yang, C., Cheung, Y.-M., Ding, J., Tan, K. C., Xue, B., and Zhang, M. Contrastive learning assisted-alignment for partial domain adaptation. IEEE Trans. Neural Netw. Learn. Syst., 34 0 (10): 0 7621--7634, Feb. 2023

    2023

  49. [57]

    Understanding deep learning (still) requires rethinking generalization

    Zhang, C., Bengio, S., Hardt, M., Recht, B., and Vinyals, O. Understanding deep learning (still) requires rethinking generalization. Commun. ACM, 64 0 (3): 0 107--115, Feb. 2021

    2021

  50. [58]

    Importance weighted adversarial nets for partial domain adaptation

    Zhang, J., Ding, Z., Li, W., and Ogunbona, P. Importance weighted adversarial nets for partial domain adaptation. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Salt Lake City, UT, USA, June 2018

    2018

Pith tools

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