Pith. sign in

REVIEW 4 major objections 5 minor 45 references

Semi-supervised representation learning via dual autoencoders for domain adaptation

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read SSRLDA claims that domain adaptation improves when a global distribution-aligned autoencoder is combined with a per-class autoencoder that aligns same-category instances, and reports state-of-the-art accuracy on four benchmarks.

desk verdict A clean closed-form dual-autoencoder idea for domain adaptation, but the headline accuracy numbers may partly reflect pseudo-label block structure rather than transferable features. read the letter →

arxiv 1908.01342 v4 pith:IDONRPCT submitted 2019-08-04 cs.LG stat.ML

Shuai Yang , Hao Wang , Yuhong Zhang , Pei-Pei Li , Yi Zhu , Xuegang Hu This is my paper · ORCID
classification cs.LGstat.ML
keywords domainadaptationsemi-supervisedrepresentationlearningdualautoencodersmarginalizeddenoisingautoencodermaximummeandiscrepancyconditionaldistributionalignmentpseudo-labelingtransfer
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

Domain adaptation tries to classify an unlabeled target domain using a labeled source domain whose distribution differs. The paper proposes SSRLDA, a representation-learning method built from two autoencoders that are trained in closed form: one learns a global representation of both domains while explicitly shrinking the marginal and class-conditional distribution gaps, and the other learns separate per-class representations using source labels and target pseudo-labels. The two representations are concatenated and a classifier is trained on the source side of that combined space. The paper reports that this design outperforms eight existing methods on four benchmarks, raising average accuracy on text tasks from 84.94% for the best baseline to 92.80% and on Office-Caltech10 from 41.23% to 49.46%. The interest is that local, category-aware alignment, not just global distribution matching, is what carries the gain, and that the whole pipeline needs no backpropagation.

What carries the argument

The object that carries the argument is the closed-form linear denoising map $W$ of a marginalized denoising autoencoder, regularized by class-conditional MMD matrices. Given corrupted copies of the input, MDAad solves $\min_W \|X - \tilde{X}W\|_F^2 + \lambda\|W\|_F^2 + \beta\, \mathrm{tr}(W^T \tilde{X}^T (\sum_{c=0}^{|C|} M_c) \tilde{X} W)$, whose optimum is $W = (E[Q]+\lambda I_d+\beta E[Q_2])^{-1}E[P]$, with $E[Q_2]$ built from the MMD matrices $M_0$ (marginal) and $M_c$ (class $c$). MMDA solves the same problem on each subset $X^{(c)}$. Because the objective is quadratic, each layer's weights are obtained in closed form, so the stack is trained greedily without backpropagation; the MMD matrices, especially the class-conditional $M_c$, are what encode the paper's claim that matching same-category instances across domains matters.

What would settle it

Measure accuracy of SSRLDA on a task where the initial SVM's target pseudo-labels are known to be poor (for example, a synthetic shift where target class proportions are inverted), and compare it against the same pipeline run with oracle target labels; if the oracle version is not clearly better, the paper's claimed mechanism, conditional distribution alignment, is not what drives the results.

Watch

Extended reading notes

Core claim

The paper's central claim is that representation learning for domain adaptation should be dual: rather than finding a single shared subspace for source and target data, learn a global representation and a local, per-class representation at the same time. The global branch (MDAad) minimizes the reconstruction error of a marginalized denoising autoencoder plus a maximum mean discrepancy (MMD) penalty that combines the marginal discrepancy between domains and the conditional discrepancy within each class; the per-class branch (MMDA) splits the data into class subsets using source labels and target pseudo-labels and runs the same closed-form denoiser on each subset. The final feature is the concatenation of the two stacks, and on that feature space a classifier is trained on source and applied to target. The paper claims this yields richer features that transfer better, and reports consistent gains over eight baselines across four datasets.

Load-bearing premise

The per-class alignment is only as good as the target pseudo-labels: the initial SVM labels and the iteratively updated classifier labels determine which target instances go into each class subset for the conditional MMD term, and if those pseudo-labels are wrong the method aligns mismatched classes.

Editorial extensions

If this is right

  • If SSRLDA is right, class-conditional alignment is buying most of the performance: the reported averages rise from 84.94% to 92.80% on text data and from 41.23% to 49.46% on Office-Caltech10 versus the strongest baseline.
  • The dual design is the reason: ablations in the paper show OMDAad or OMMDA alone is weaker than the combination, so both global and per-class views carry information.
  • Hierarchical or multi-class text datasets gain more from the local branch, suggesting per-class alignment is especially valuable when classes have internal structure.
  • Closed-form training means the method can be applied to high-dimensional text features without a GPU or end-to-end training, which matters for practical deployment.
  • The generalization bound supplied in the paper implies that target error is controlled by source empirical error plus the MMD terms that the two autoencoders explicitly minimize, giving a principled reason for the design.

Reading between the lines

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

  • A natural stress test the paper does not run is to replace the target pseudo-labels with random class assignments in the $M_c$ matrices; if the gains persist, the class-conditional term is not actually the mechanism, and if they vanish, pseudo-label accuracy is the real lever.
  • The same dual global-plus-class alignment could be inserted into a deep neural encoder, at the cost of the closed-form solution; the paper's experiments do not show whether the benefit survives stochastic training.
  • Because the concatenation weights global and local features equally, an adaptive weighting, the paper's own stated future direction, could plausibly raise accuracy further on datasets where one branch dominates, such as Spam and Office-Caltech10.
  • The reliance on source labels and target pseudo-labels makes the method a natural candidate for partial domain adaptation or class-imbalanced settings, where per-class partition sizes vary; the paper does not test those regimes.
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 SSRLDA, a semi-supervised representation-learning framework for domain adaptation that combines two autoencoder variants: MDAad, which minimizes reconstruction error plus marginal and conditional maximum mean discrepancy (MMD), and MMDA, which learns per-class (local) representations using source labels and target pseudo-labels. The authors derive closed-form solutions for the linear mappings, present a domain-adaptation generalization bound, and evaluate the method on 20 Newsgroups, Reuters, Spam, and Office-Caltech10, reporting accuracy gains over eight baselines.

Significance. The dual global/local representation idea is a plausible contribution to feature-based domain adaptation, and the manuscript's algebraic derivation of the closed-form mappings (Eqs. (12) and (16)) is mostly coherent, modulo the expectation issue noted below. The paper also provides a public code repository and a standard PAC-style generalization bound. If the empirical claims hold, the approach would be a useful addition to the autoencoder-based DA literature. However, the current experimental evidence is weakened by single-run results, test-set hyperparameter selection, and the unablated pseudo-label channel in MMDA.

major comments (4)
  1. [§4.3, Eq. (13) and §4.4, Eq. (18)] The expectation formulas for E[Q2] and E[Q2^(c)] are incorrect when the MMD matrix C (or M_0^(c)) is not diagonal. For diagonal entries, E[(~X^T C ~X)_ii] = (1-p) * sum_k C_kk x_{k,i}^2 + (1-p)^2 * sum_{k≠l} C_kl x_{k,i} x_{l,i}, which is not equal to U2_ii (1-p) as stated. Since the MMD matrices in Eqs. (7) and (9) are generally non-diagonal, the closed-form W in Eqs. (12) and (16) is not the exact minimizer of the stated objectives. The derivation should be corrected, or the approximation should be explicitly identified and justified.
  2. [§5.4 and Tables 3–4] The hyperparameters (λ, β, p, l) appear to be selected by evaluating accuracy on the target test domain (the sensitivity plots in Figs. 2–4), and Tables 3–4 report a single run with no variance or statistical significance tests. This makes the claimed improvements, e.g., 92.80% vs. 84.94% average on textual datasets, difficult to assess. Please report means and standard deviations over multiple runs and use a validation split or a nested tuning procedure so that the reported numbers are not cherry-picked on the test set.
  3. [§4.4 and Algorithm 1 (lines 14–21)] The MMDA local representation encodes the target pseudo-label as a block index: each target instance is transformed by its pseudo-class-specific W^(c) and placed into block c of H2, with the other blocks zero. A classifier trained on source H2 can therefore read the pseudo-label from the block structure alone. Consequently, the reported accuracy partly measures pseudo-label accuracy rather than transferable representation structure. The conclusion acknowledges susceptibility to pseudo-label uncertainty, but the paper never measures pseudo-label accuracy, never varies pseudo-label quality, and never ablates the block-encoding mechanism (e.g., with random pseudo-labels). Please add such an analysis to support the claim that the gains come from the learned representation.
  4. [§5.2, baselines] The comparison omits closely related autoencoder-based DA methods, including SEAE [26] and SRAAR [27], which are cited in the introduction from the same group, as well as conditional-MMD baselines such as JAN. Because the central claim is state-of-the-art performance, the baseline list should include these closely related approaches or explain why they are excluded.
minor comments (5)
  1. [§4.4, Eq. (15)] The derivative expression contains an extra parenthesis in the term '2β~X(c)T M(c)0 X(c))~X(c)W(c)' that makes the equation unreadable; please fix the typo.
  2. [§4.5, Theorem 1] The quantity Θ is introduced in Eq. (22) but only defined in the subsequent proof; also, the bound is a standard domain-adaptation result and is not used to derive any algorithmic consequence. Consider stating it as a proposition and clarifying its role.
  3. [§5.6] The text initially says that 'the proxy-A-distance increases after representations learning, which indicates that the new representations are helpful', but then reports decreases on some datasets; this is contradictory and should be rephrased to explain what the proxy-A-distance can and cannot indicate.
  4. [Throughout] There are numerous typos and grammatical issues (e.g., 'antoencoders', 'different', 'staked layers', 'hyper-paremeter') that should be cleaned up during revision.
  5. [§5.4, Fig. 3] The sensitivity analysis shows that SSRLDA is 'sensitive to p' and 'sensitive to β', which undermines the claim in Section 5.4 that SSRLDA 'can achieve an optimal performance under wide range of parameter values'; please reconcile these statements.

Circularity Check

0 steps flagged · score 0.0 of 10

No load-bearing circularity identified: the autoencoder maps are closed-form solutions of explicit objectives and the accuracy comparisons are measured against external baselines.

full rationale

The derivation chain is self-contained. MDAad's W in Eq. (12) and MMDA's W(c) in Eq. (16) are closed-form solutions of stated regularized objectives; the MMD matrices Mc in Eq. (9) are built from source labels and target pseudo-labels, which are predictions, not ground-truth labels used in the evaluation. The empirical comparisons in Tables 3-4 are against external baselines and the accuracy numbers are measured on target true labels, not read back from fitted constants. The only self-referential aspect is the iterative pseudo-label loop (Algorithm 1 lines 1 and 23), and the Conclusion explicitly acknowledges susceptibility to pseudo-label accuracy; this is a robustness/self-training concern, not a tautological derivation, since the target pseudo-labels are not the optimization target and can be wrong. The self-citations [26] and [27] appear in the related-work survey and are not load-bearing for the proposed objective or bounds. Theorem 1's bound follows from external domain-adaptation theory [42,43,45]; the paper does not invoke a self-authored uniqueness theorem. Hence no circular step is exhibited.

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

The method's derivations rest on standard MDA expectation formulas, the domain adaptation bound of Ben-David, the Lemma from Ghifary et al., and a set of per-dataset hyperparameters chosen from test accuracy. No new physical or conceptual entities are introduced; MDAad and MMDA are algorithmic variants.

free parameters (5)
  • lambda (reconstruction regularization weight) = 10 (Spam, Reuters), 1E-5 (20 Newsgroups, Office-Caltech)
    Chosen per dataset in Section 5.4 to maximize test accuracy, not on a held-out validation split.
  • beta (MMD regularization weight) = 0.1 (Spam, Reuters), 1000 (20 Newsgroups), 1 (Office-Caltech)
    Selected from accuracy curves in Figure 4; controls the strength of distribution alignment.
  • p (noise probability) = 0.9 (20 Newsgroups, Spam), 0.6 (Office-Caltech)
    Selected from Figure 3; the authors state SSRLDA is sensitive to p.
  • l (number of stacked layers) = 4 (20 Newsgroups, Reuters, Spam), 3 (Office-Caltech)
    Selected from Figure 2 based on accuracy and training time.
  • SVM pseudo-label classifier hyperparameters = not specified
    Pseudo-labels are produced by an SVM in Section 4.3, but its parameters are not reported, even though pseudo-label quality drives the per-class partitions used by MMDA and MDAad.
assumptions (5)
  • domain assumption Source and target domains share the same label space with the same |C| classes.
    Required for per-class partitions X_S^(c), X_T^(c) and conditional MMD in Eqs. (8)-(9) and Algorithm 1.
  • standard math The marginalized denoising autoencoder closed-form solution and expectation formulas from Chen et al. [11] are correct and applicable to the MMD-augmented objective.
    The paper builds directly on the MDA objective and expectation limits without re-deriving them.
  • standard math The domain adaptation generalization bound in Theorem 1 and Lemma 1 from Ghifary et al. [45] hold as stated and apply to the MMD terms used.
    The proof reuses standard Ben-David bounds and Lemma 1 is cited from [45]; no new derivation is provided.
  • domain assumption The tanh nonlinearity and layer-wise greedy stacking capture transferable nonlinear structure.
    Chosen following prior autoencoder DA work; no theoretical support specific to SSRLDA is given.
  • domain assumption Pseudo-labels obtained from a source-trained SVM are accurate enough that per-class MMD terms improve rather than harm alignment.
    Load-bearing for Eqs. (8) and (14); the paper acknowledges susceptibility to pseudo-label uncertainty in the Conclusion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Semi-supervised representation learning via dual autoencoders for domain adaptation." pith.science (2026). https://pith.science/paper/IDONRPCT

@misc{pith2026190801342,
  author       = {Pith},
  title        = {Pith review of: Semi-supervised representation learning via dual autoencoders for domain adaptation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IDONRPCT}},
  note         = {Machine review of arXiv:1908.01342}
}
read the original abstract

Domain adaptation aims to exploit the knowledge in source domain to promote the learning tasks in target domain, which plays a critical role in real-world applications. Recently, lots of deep learning approaches based on autoencoders have achieved a significance performance in domain adaptation. However, most existing methods focus on minimizing the distribution divergence by putting the source and target data together to learn global feature representations, while they do not consider the local relationship between instances in the same category from different domains. To address this problem, we propose a novel Semi-Supervised Representation Learning framework via Dual Autoencoders for domain adaptation, named SSRLDA. More specifically, we extract richer feature representations by learning the global and local feature representations simultaneously using two novel autoencoders, which are referred to as marginalized denoising autoencoder with adaptation distribution (MDAad) and multi-class marginalized denoising autoencoder (MMDA) respectively. Meanwhile, we make full use of label information to optimize feature representations. Experimental results show that our proposed approach outperforms several state-of-the-art baseline methods.

Figures

Figures reproduced from arXiv: 1908.01342 by the authors.

Figure 1
Figure 1. The whole framework of our proposed SSRLDA [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Accuracy (%) with different number of layers l on different datasets. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Accuracy (%) with different values of p on different datasets. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Accuracy (%) with different values of β on different datasets. 16 [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Comparison of OMDAad, OMMDA and SSRLDA on different datasets. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: The Proxy-A-distance on raw data and SSRLDA. [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: The Proxy-A-distance on mSDA and SSRLDA. [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages

  1. [26]

    S. Yang, Y . Zhang, Y . Zhu, P. Li, X. Hu, Representation learning via serial autoencoders for domain adaptation, Neurocomputing 351 (2019) 1–9

    work page 2019

  2. [27]

    S. Yang, Y . Zhang, Y . Zhu, H. Ding, X. Hu, The l2,1-norm stacked robust autoencoders via adaptation regularization for domain adaptation, in: Artificial Intelligence and Security - 5th International Conference, ICAIS, July 26-28, New York, NY , USA, 2019, pp. 100–111

    work page 2019

  3. [1]

    H. D. III, Frustratingly easy domain adaptation, in: Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics, June 23-30, Prague, Czech Republic, 2007, pp. 256–263

    work page 2007

  4. [2]

    P. Wei, Y . Ke, C. Goh, Domain specific feature transfer for hybrid domain adaptation, in: IEEE International Conference on Data Mining, November 18-21, New Orleans, LA, USA, 2017, pp. 1027–1032

    work page 2017

  5. [3]

    Pouyanfar, S

    S. Pouyanfar, S. Sadiq, Y . Yan, H. Tian, Y . Tao, M. E. P. Reyes, M. Shyu, S. Chen, S. S. Iyengar, A survey on deep learning: Algorithms, techniques, and applications, ACM Comput. Surv. 51 (5) (2019) 92:1–92:36

    work page 2019

  6. [4]

    W. Yang, W. Lu, V . Zheng, A simple regularization-based algorithm for learning cross-domain word embeddings, in: Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, September 9-11, Copenhagen, Denmark, 2017, pp. 2898–2904

    work page 2017

  7. [5]

    Jiang, H

    W. Jiang, H. Gao, W. Lu, W. Liu, F. Chung, H. Huang, Stacked robust adaptively regularized auto-regressions for domain adaptation, IEEE Trans. Knowl. Data Eng. 31 (3) (2019) 561–574

    work page 2019

  8. [6]

    S. J. Pan, Q. Yang, A survey on transfer learning, IEEE Trans. Knowl. Data Eng. 22 (10) (2010) 1345–1359

    work page 2010

Show all 45 references
  1. [7]

    Sagha, N

    H. Sagha, N. Cummins, B. Schuller, Stacked denoising autoencoders for sentiment analysis: a review, Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery 7 (5) (2017) e1212

    2017

  2. [8]

    C. Tan, F. Sun, T. Kong, W. Zhang, C. Yang, C. Liu, A survey on deep transfer learning, in: 27th International Conference on Artificial Neural Networks, October 4-7, Rhodes, Greece, 2018, pp. 270–279

    2018

  3. [9]

    Zellinger, B

    W. Zellinger, B. A. Moser, T. Grubinger, E. Lughofer, T. Natschl ¨ager, S. Saminger-Platz, Robust unsupervised domain adaptation for neural networks via moment alignment, Inf. Sci. 483 (2019) 174–191

    2019

  4. [10]

    Glorot, A

    X. Glorot, A. Bordes, Y . Bengio, Domain adaptation for large-scale sentiment classification: A deep learning approach, in: Proceedings of the 28th International Conference on Machine Learning, June 28 - July 2, Bellevue, Washington, USA, 2011, pp. 513–520

    2011

  5. [11]

    M. Chen, Z. E. Xu, K. Q. Weinberger, F. Sha, Marginalized denoising autoencoders for domain adaptation, in: Proceedings of the 29th International Conference on Machine Learning, June 26-July 1, Edinburgh, Scotland, UK, 2012, pp. 767–774

    2012

  6. [12]

    L. Meng, S. Ding, N. Zhang, J. Zhang, Research of stacked denoising sparse autoencoder, Neural Computing and Applications 30 (7) (2018) 2083–2100

    2018

  7. [13]

    P. Wei, Y . Ke, C. K. Goh, Feature analysis of marginalized stacked denoising autoenconder for unsupervised domain adaptation, IEEE Transactions on Neural Networks and Learning Systems PP (99) (2018) 1–14

    2018

  8. [14]

    B. Du, W. Xiong, J. Wu, L. Zhang, L. Zhang, D. Tao, Stacked convolutional denoising auto-encoders for feature representation, IEEE Trans. Cybernetics 47 (4) (2017) 1017–1027

    2017

  9. [15]

    F. I. Eyiokur, D. Yaman, H. K. Ekenel, Domain adaptation for ear recognition using deep convolutional neural networks, IET Biometrics 7 (3) (2018) 199–206

    2018

  10. [16]

    Z. Gao, C. Shen, C. Xie, Stacked convolutional auto-encoders for single space target image blind deconvolution, Neurocomputing 313 (2018) 295–305

    2018

  11. [17]

    Jaech, L

    A. Jaech, L. P. Heck, M. Ostendorf, Domain adaptation of recurrent neural networks for natural language understanding, in: 17th Annual Conference of the International Speech Communication Association, September 8-12, San Francisco, CA, USA, 2016, pp. 690–694

    2016

  12. [18]

    Y . Ding, J. Yu, J. Jiang, Recurrent neural networks with auxiliary labels for cross-domain opinion target extraction, in: Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence, February 4-9, San Francisco, California, USA, 2017, pp. 3436–3442

    2017

  13. [19]

    H. Deng, L. Zhang, X. Shu, Feature memory-based deep recurrent neural network for language modeling, Appl. Soft Comput. 68 (2018) 432–446

    2018

  14. [20]

    Tzeng, J

    E. Tzeng, J. Ho ffman, K. Saenko, T. Darrell, Adversarial discriminative domain adaptation, in: IEEE Conference on Computer Vision and Pattern Recognition, July 21-26, Honolulu, HI, USA, 2017, pp. 2962–2971

    2017

  15. [21]

    M. Long, Z. Cao, J. Wang, M. I. Jordan, Conditional adversarial domain adaptation, in: Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems, NeurIPS, December 3-8, Montr´eal, Canada., 2018, pp. 1647–1657

    2018

  16. [22]

    Z. Cao, L. Ma, M. Long, J. Wang, Partial adversarial domain adaptation, in: Computer Vision-ECCV-15th European Conference, September 8-14, Munich, Germany, 2018, pp. 139–155

    2018

  17. [23]

    J. Shen, Y . Qu, W. Zhang, Y . Yu, Wasserstein distance guided representation learning for domain adaptation, in: Proceedings of the Thirty- Second AAAI Conference on Artificial Intelligence, (AAAI-18), February 2-7, New Orleans, Louisiana, USA, 2018, pp. 4058–4065

    2018

  18. [24]

    P. Wei, Y . Ke, C. K. Goh, Deep nonlinear feature coding for unsupervised domain adaptation, in: Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence, IJCAI, July 9-15, New York, NY , USA, 2016, pp. 2189–2195

    2016

  19. [25]

    Csurka, B

    G. Csurka, B. Chidlovskii, S. Clinchant, S. Michel, An extended framework for marginalized domain adaptation, CoRR abs /1702.05993

    arXiv

  20. [28]

    Blitzer, R

    J. Blitzer, R. T. McDonald, F. Pereira, Domain adaptation with structural correspondence learning, in: EMNLP, Proceedings of the 2006 Conference on Empirical Methods in Natural Language Processing, July 22-23, Sydney, Australia, 2006, pp. 120–128

    2006

  21. [29]

    S. J. Pan, I. W. Tsang, J. T. Kwok, Q. Yang, Domain adaptation via transfer component analysis, IEEE Trans. Neural Networks 22 (2) (2011) 199–210

    2011

  22. [30]

    J. Li, Y . Wu, K. Lu, Structured domain adaptation, IEEE Trans. Circuits Syst. Video Techn. 27 (8) (2017) 1700–1713

    2017

  23. [31]

    B. Sun, J. Feng, K. Saenko, Return of frustratingly easy domain adaptation, in: Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence, February 12-17, Phoenix, Arizona, USA., 2016, pp. 2058–2065

    2016

  24. [32]

    Sharma, P

    R. Sharma, P. Bhattacharyya, S. Dandapat, H. S. Bhatt, Identifying transferable information across domains for cross-domain sentiment 21 classification, in: Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics, Melbourne, Australia, July 15-20...

    2018

  25. [33]

    S. Di, J. Peng, Y . Shen, L. Chen, Transfer learning via feature isomorphism discovery, in: Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD, August 19-23, London, UK, 2018, pp. 1301–1309

    2018

  26. [34]

    Y . Luo, Y . Wen, T. Liu, D. Tao, Transferring knowledge fragments for learning distance metric from a heterogeneous domain, IEEE Trans. Pattern Anal. Mach. Intell. 41 (4) (2019) 1013–1026

    2019

  27. [35]

    Y . Chen, S. Song, S. Li, L. Yang, C. Wu, Domain space transfer extreme learning machine for domain adaptation, IEEE Trans. Cybernetics 49 (5) (2019) 1909–1922

    2019

  28. [36]

    Y . He, Y . Tian, D. Liu, Multi-view transfer learning with privileged learning framework, Neurocomputing 335 (2019) 131–142

    2019

  29. [37]

    F. Liu, J. Lu, G. Zhang, Unsupervised heterogeneous domain adaptation via shared fuzzy equivalence relations, IEEE Trans. Fuzzy Systems 26 (6) (2018) 3555–3568

    2018

  30. [38]

    Jiang, H

    W. Jiang, H. Gao, F. Chung, H. Huang, The l2, 1-norm stacked robust autoencoders for domain adaptation, in: Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence, February 12-17, Phoenix, Arizona, USA., 2016, pp. 1723–1729

    2016

  31. [39]

    Ganin, V

    Y . Ganin, V . S. Lempitsky, Unsupervised domain adaptation by backpropagation, in: Proceedings of the 32nd International Conference on Machine Learning, ICML, July 6-11, Lille, France, 2015, pp. 1180–1189

    2015

  32. [40]

    Clinchant, G

    S. Clinchant, G. Csurka, B. Chidlovskii, A domain adaptation regularization for denoising autoencoders, in: Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, ACL, August 7-12, Berlin, Germany, V olume 2: Short Papers, 2016, pp. 26–31

    2016

  33. [41]

    Y . Zhu, X. Hu, Y . Zhang, P. Li, Transfer learning with stacked reconstruction independent component analysis, Knowl.-Based Syst. 152 (2018) 100–106

    2018

  34. [42]

    Ben-David, J

    S. Ben-David, J. Blitzer, K. Crammer, F. Pereira, Analysis of representations for domain adaptation, in: Advances in Neural Information Pro- cessing Systems 19, Proceedings of the Twentieth Annual Conference on Neural Information Processing Systems, December 4-7, Vancouver, Br...

    2006

  35. [43]

    Mansour, M

    Y . Mansour, M. Mohri, A. Rostamizadeh, Domain adaptation: Learning bounds and algorithms, in: The 22nd Conference on Learning Theory, June 18-21, Montreal, Quebec, Canada, 2009, pp. 100–111

    2009

  36. [44]

    Vapnik, Statistical learning theory, Wiley, 1998

    V . Vapnik, Statistical learning theory, Wiley, 1998

    1998

  37. [45]

    Ghifary, D

    M. Ghifary, D. Balduzzi, W. B. Kleijn, M. Zhang, Scatter component analysis: A unified framework for domain adaptation and domain generalization, IEEE Trans. Pattern Anal. Mach. Intell. 39 (7) (2017) 1414–1430. 22

    2017

Pith tools

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