Pith. sign in

REVIEW 4 major objections 4 minor 37 references

Towards a Problem-Oriented Domain Adaptation Framework for Machine Learning

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper argues that any single-source, homogeneous, unsupervised domain adaptation problem can be classified into one of five scenarios, each with its own solution family.

desk verdict A useful and honest packaging of the Moreno-Torres shift taxonomy into a practitioner-oriented framework, but the abstract overclaims what the framework can actually do a priori, and the evaluation is thinner than the claims. read the letter →

arxiv 2501.04528 v1 pith:KKZW6L5V submitted 2025-01-08 cs.LG cs.AI

Philipp Spitzer , Dominik Martin , Laurin Eichberger , Niklas Kühl This is my paper · ORCID
classification cs.LGcs.AI
keywords domainadaptationshiftdatasettransferlearningcausalityproblem-orientedframeworkdecisionsupportunsupervised
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 tries to make domain adaptation usable by people who are not specialists in it. Its central proposal is a framework that classifies every single-source, homogeneous, unsupervised domain adaptation problem into one of five scenarios—prior shift, class-conditional shift, covariate shift, concept shift, and general data set shift—based on the causal direction between features and labels and on which probability distributions change between source and target. For each scenario the framework names a family of solution approaches, from class-based reweighting and transformation learning to instance-based reweighting, and it provides procedures for deciding which scenario a problem belongs to. The authors claim the framework has the explanatory power to capture any domain adaptation problem effectively, and they support this with simulations, three real-world case studies, and a 100-participant experiment in which decision support markedly improved scenario identification. The paper also concedes that expert knowledge remains decisive for scenario identification and that a priori guarantees of adaptation success are out of reach.

What carries the argument

The machinery is a two-step classification rule plus a set of importance-weighting identities. First, determine the causal direction between $X$ and $Y$; this splits the five scenarios into two groups. Second, compare which distributions differ, using the joint-probability decomposition above. For prior shift and covariate shift, the target risk can be rewritten as a source risk with weights—$\hat{R}_t(h)=\frac{1}{n}\sum_i \ell(h(x_i),y_i)w(y_i)$ for prior shift, and an analogous feature-space weight for covariate shift—so the recommended solutions are weighting schemes. For class-conditional shift, the defining assumption is that a transformation $t$ exists with $P_s(x)=P_t(t(x))$; the recommended solutions are accordingly transformation-learning methods, either direct domain mapping or mapping both domains into a shared invariant space. The general shift scenario inherits the transformation-learning family but is flagged by bounds showing that no guarantee survives a large prior difference.

What would settle it

Take a set of domain adaptation benchmarks with artificially constructed ground-truth shifts (each of the five types, in both causal directions). Have users with no prior domain adaptation expertise run the paper's determination procedures on each benchmark and then implement the recommended family. The central claim fails if a substantial fraction of causal attributions is wrong, or if the recommended family does not beat the other families on average. A cheaper variant: rerun the 100-participant study on deliberately causality-ambiguous cases; if performance with the framework no longer exceeds the control group, the framework's guidance is not doing the work.

Watch

Extended reading notes

Core claim

The central discovery, stated on the paper's own terms, is that the space of domain adaptation methods becomes navigable once a problem is located on two axes: causal direction ($X\to Y$ when features cause the label, $Y\to X$ when the label causes the features) and which factor in the decomposition $P(x,y)=P(y\mid x)P(x)=P(x\mid y)P(y)$ changes between domains. Under $Y\to X$, a change in $P(y)$ with stable $P(x\mid y)$ is a prior shift, whereas changed class conditionals $P(x\mid y)$ with stable priors form a class-conditional shift. Under $X\to Y$, a change in $P(x)$ with stable concept $P(y\mid x)$ is a covariate shift, and a changed concept with stable features is a concept shift—treated in the framework as not a domain adaptation problem. Combined changes fall into a general data set shift, for which the framework recommends the class-conditional-shift toolbox with no performance guarantee. Each scenario carries a specific solution family, and the evaluation episodes are presented as evidence that the scenario definitions are internally consistent, explain observed performances on known benchmarks, and improve nonexperts' ability to identify shifts.

Load-bearing premise

The load-bearing premise is that a practitioner can correctly determine whether features cause labels or labels cause features ($X\to Y$ vs $Y\to X$) before applying the framework; if that attribution is wrong, the framework selects the wrong scenario and recommends the wrong solution family.

Editorial extensions

If this is right

  • A practitioner who can name the causal direction and one or two changed distributions can choose a solution family without surveying the domain adaptation literature.
  • Misclassification is not harmless: the paper reports that treating a class-conditional shift as a covariate shift can make performance worse, so the taxonomy doubles as a diagnostic for failed adaptation attempts.
  • Under covariate shift, domain adaptation is only needed when the model is misspecified; a well-specified model trained on enough source samples should already generalize, which reframes 'when to adapt' as a model-capacity question.
  • For general data set shifts no method comes with a performance guarantee, so the framework's honest output is often a solution family to try, not a method to trust.
  • The 100-participant experiment suggests the taxonomy is teachable: participants with framework support identified shifts far more often than those without, across covariate, prior, and class-conditional cases.

Reading between the lines

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

  • The framework's dependence on causal direction points to a natural extension the paper only gestures at: for causality-ambiguous problems, causal discovery algorithms could be run first and their output fed into the scenario tables; the paper currently leaves that as an open requirement.
  • Because the paper finds that within a solution family the best implementation is problem-dependent, a testable benchmark could pit 'random method from the recommended family' against 'random method from a wrong family' to quantify the framework's practical ceiling.
  • The stated scope (single-source, homogeneous, unsupervised) suggests the taxonomy has not yet been shown to extend to multi-source or heterogeneous settings; a plausible but unproven generalization is that the same causality-first logic would still partition those problems.
  • The evaluation measures whether users can identify shifts, not whether following the recommendations improves final target accuracy; a direct end-to-end randomized study would settle how much of the framework's benefit survives actual deployment.
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 a problem-oriented domain adaptation framework developed through design science research. It classifies single-source, homogeneous, unsupervised domain adaptation problems into five scenarios (prior shift, class-conditional shift, covariate shift, concept shift, and general dataset shift) using Moreno-Torres' dataset shift taxonomy and the causal direction between features X and labels Y. For each scenario, Table 2 recommends a family of solution procedures, and Table 3 proposes procedures for scenario determination. The framework is evaluated in three episodes: synthetic experiments (EE1), literature examples (EE2), and a 100-participant between-subject study (EE3). The paper claims that the evaluation demonstrates the framework's explanatory power to capture any domain adaptation problem and that it provides clear guidance for practitioners lacking in-depth domain adaptation knowledge.

Significance. If the claims were fully supported, the framework would be a practically useful resource: it condenses established dataset-shift definitions into a concise taxonomy, links each scenario to solution families, and explicitly foregrounds causal direction, which is often neglected. The DSR methodology, the summary tables, and the multi-episode evaluation effort are strengths, as is the candid discussion of no-free-lunch and of the framework's limitations. However, the contribution as currently framed is a classification aid and ex-post interpretive tool rather than an a priori suitability test; the central claims exceed what the evaluation and the paper's own Discussion establish. The manuscript therefore needs substantial revision to align its claims with its evidence, but the underlying taxonomy and solution mapping are defensible and potentially useful.

major comments (4)
  1. [Section 6 / Section 1.2] The Discussion states: 'In conclusion, we could not find a way to determine the suitability of domain adaptation for any given problem (by indicators that are available a priori).' This directly contradicts RQ1 and design requirement (2) in Section 1.2, which promise that practitioners can 'a priori and with reasonable effort determine the correct scenario.' It also undercuts the Abstract's claim to provide 'clear guidance' and 'explanatory power to capture any domain adaptation problem effectively.' The paper should either remove the a priori suitability claim and reframe the framework as an ex-post interpretive aid, or supply a genuine a priori determination procedure. As written, the central claim is not supported by the framework's own conclusion.
  2. [Table 3 / Section 4.6] Three rows of Table 3 (Prior, Class-cond., Covariate) list 'application of solution procedure with a beneficial outcome' as a scenario-determination criterion. This is not an a priori test: it requires running a candidate solution before deciding which scenario is present, and Section 6's caveat that a failed domain adaptation attempt does not prove impossibility applies symmetrically, so a beneficial outcome does not uniquely identify the scenario either (no-free-lunch, Section 4.4). Within the paper's own unsupervised, no-target-labels scope, the remaining non-circular criteria for class-conditional and general shifts largely reduce to domain knowledge, which the Discussion concedes cannot always be statistically verified. Table 3 therefore does not provide the decision support promised to practitioners who lack domain expertise.
  3. [Section 5.2.1 / Table 6] The heart disease evaluation reports that MMD kernel matching with an RBF kernel improves accuracy from 0.47 to 0.54 (Section 5.2.1), but no error bars, number of runs, significance test, or sensitivity analysis for the RBF bandwidth are reported, and no code or data are provided. Similarly, Table 6 reports improvements such as 91.3 to 92.0 without uncertainty. This single point estimate is too thin to support the 'Empirical Validation' contribution in Section 1.3 or the assertion in Section 6 that following the framework 'delivers significant benefit over an uninformed domain adaptation attempt.'
  4. [Sections 5.2, 5.3, 6] Evaluation Episode 3 measures participants' ability to identify the correct scenario, not whether following the framework improves the outcome of an actual domain adaptation attempt; the chi-squared test supports only the former. Section 6 nevertheless concludes that the framework 'delivers significant benefit over an uninformed domain adaptation attempt,' which the experimental design does not establish. In addition, Evaluation Episode 2 classifies literature examples into the framework's own categories and then uses the known outcomes of those examples as confirmation, which is partly circular as evidence for the framework's 'explanatory power.'
minor comments (4)
  1. [Throughout] The manuscript contains unresolved '??' placeholders for cross-references and equations in Sections 1.2, 4.3, 4.4, and 5.1; these must be fixed before publication.
  2. [Figure 12] The legend for the concept-shift panel lists 'Pt(y = +1|x)' twice; the second entry should presumably be 'Pt(y = -1|x)'.
  3. [Section 5.2.1] The sentence 'However, the authors do not publish the exact methods and results' is ambiguous because 'the authors' could refer to Kouw and Loog or to the current paper's authors; please clarify.
  4. [Section 1.3 / Section 6] Section 1.3 claims the framework has been 'rigorously evaluated across multiple datasets,' while Section 6 acknowledges that 'a larger-scale summative evaluation is still needed'; the wording in Section 1.3 should be moderated to match this limitation.

Circularity Check

1 steps flagged · score 6.0 of 10

Table 3's 'beneficial outcome' criterion makes scenario identification depend on the success of the very solution procedure the scenario is supposed to select; Section 6 then concedes no a priori suitability test was found.

  1. self definitional [Section 4.6, Table 3 (Prior row; identical 'beneficial outcome' clauses in Class-cond. and Covariate rows)]
    "Determine shift in priors via domain knowledge, semi-supervised hypothesis testing (e.g., Kolmogorov-Smirnoff), maximum-likelihood hypothesis testing, e.g., Saerens et al. (2002), or application of solution procedure with a beneficial outcome."

    Table 3 is the operational answer to RQ1 for determining which scenario applies. Its Prior, Class-cond., and Covariate rows accept 'application of solution procedure with a beneficial outcome' as a determination criterion. But Table 2 prescribes the solution procedure only after the scenario is known, so success of the recommended procedure is used both to infer the scenario and to justify the recommendation. This is a closed loop: the outcome being predicted (appropriateness of a DA approach) is the indicator for the scenario that generates the prediction.

full rationale

The scenario definitions and solution families are not themselves circular: they are taken from Moreno-Torres (2012), Kouw and Loog (2021), and Wilson and Cook (2020), and the mapping from distributional shift types to solution families is an external literature-based construction. There is no self-citation chain or imported uniqueness theorem. The circularity is localized in the determination procedure of Table 3, where 'application of solution procedure with a beneficial outcome' uses the success of a Table 2 solution as evidence for the scenario that selects that solution. This matters because RQ1 explicitly asks for an a priori determination; Section 6 openly disowns that goal: 'we could not find a way to determine the suitability of domain adaptation for any given problem (by indicators that are available a priori)'. Evaluation Episode 2 re-describes known literature results through the framework and cites their prior empirical outcomes, so it is a consistency check rather than an independent confirmation of explanatory power; Episode 3 tests whether participants can reproduce the authors' own scenario labels, not whether those labels are externally grounded. These facts support a partial circularity score: the central a priori-guidance claim is not derivable without running the solution, but the framework's taxonomy and per-scenario recommendations retain independent content from the cited literature.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The framework itself is a conceptual artifact, not a physical entity. The load-bearing assumptions are the completeness of the Moreno-Torres taxonomy, the knowability of causal direction, the existence of a transformation function for class-conditional shift, and the correctness of quoted error bounds. The only apparent free parameter is the unreported kernel bandwidth in the empirical reweighting example.

free parameters (1)
  • RBF kernel bandwidth for MMD reweighting in heart disease experiment = not reported
    The weighted accuracy of 0.54 reported in Section 5.2.1 likely depends on this hyperparameter, but the value is not given, so the result cannot be independently tuned or reproduced.
assumptions (4)
  • domain assumption The dataset shift taxonomy of Moreno-Torres (2012) with five shift types is a complete description of single-source homogeneous unsupervised domain adaptation problems.
    The framework's exhaustive 'General' category is adopted from Moreno-Torres without independent justification; Section 4 states scenarios are 'the result of applying Moreno-Torres' unified view on data shifts.'
  • domain assumption The causal direction of the data-generating process (X to Y or Y to X) is knowable before solving the problem.
    Section 4.6 says causality must be determined by domain knowledge and, if unclear, causal research must be performed; the framework's scenario selection depends on this being available.
  • domain assumption For class-conditional shift, a transformation function t exists such that Ps(x) = Pt(t(x)).
    Section 4.4 asserts this relationship and defers the proof to a '??' placeholder; this existence is required for the recommended feature-space alignment solutions.
  • standard math The error bounds quoted from Cortes et al. (2010), Gong et al. (2016) and Zhao et al. (2019) are correct and apply to the stated scenarios.
    Equations 4, 5, 6 and 7 are used to justify the scenario-specific recommendations; these are external results adopted without independent verification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards a Problem-Oriented Domain Adaptation Framework for Machine Learning." pith.science (2026). https://pith.science/paper/KKZW6L5V

@misc{pith2026250104528,
  author       = {Pith},
  title        = {Pith review of: Towards a Problem-Oriented Domain Adaptation Framework for Machine Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KKZW6L5V}},
  note         = {Machine review of arXiv:2501.04528}
}
read the original abstract

Domain adaptation is a sub-field of machine learning that involves transferring knowledge from a source domain to perform the same task in the target domain. It is a typical challenge in machine learning that arises, e.g., when data is obtained from various sources or when using a data basis that changes over time. Recent advances in the field offer promising methods, but it is still challenging for researchers and practitioners to determine if domain adaptation is suitable for a given problem -- and, subsequently, to select the appropriate approach. This article employs design science research to develop a problem-oriented framework for domain adaptation, which is matured in three evaluation episodes. We describe a framework that distinguishes between five domain adaptation scenarios, provides recommendations for addressing each scenario, and offers guidelines for determining if a problem falls into one of these scenarios. During the multiple evaluation episodes, the framework is tested on artificial and real-world datasets and an experimental study involving 100 participants. The evaluation demonstrates that the framework has the explanatory power to capture any domain adaptation problem effectively. In summary, we provide clear guidance for researchers and practitioners who want to employ domain adaptation but lack in-depth knowledge of the possibilities.

Figures

Figures reproduced from arXiv: 2501.04528 by the authors.

Figure 1
Figure 1. The number of search results for domain adaptation in Google Scholar by year indicates a growing interest in the topic. very few labeled training data available, which in turn limits the usefulness of supervised learning approaches. Creating labeled examples (e.g., through manually labeling data) is often expensive or not even possible to perform on the required scale. For example, in the well-known Cityscapes datas… view at source ↗
Figure 2
Figure 2. Domain adaptation results for CyCada by Ho [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Answer rates for questions concerning domain adaptation on the StackExchange network are significantly below the average of 70%. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (17 more)
Figure 4
Figure 4. Figure 4: The road map illustrates how the artifact, i.e., the framework, is constructed and improved by the two interconnected steps: theory [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: In Y → X systems, a prior shift occurs if the marginal label distributions (left) change, but the class conditionals (right) remain the 1 same. 13 [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: In X → Y systems, a covariate shift occurs if the distribution in the feature space (left) changes, but the concept (right) remains the same. 4.2. Covariate Shift A covariate shift appears only in X → Y problems and is defined as the case where Ps(y|x) = Pt(y|x) and Ps…
Figure 7
Figure 7. Figure 7: In X → Y systems, a concept shift is the opposite of a covariate shift: The feature distribution (left) remains constant, but the task (right) changes in the target domain. y=+1 y = −1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 P(y) Source Target Source Target −5.0 −2.5 0.0 2.5 5.0 0…
Figure 8
Figure 8. Figure 8: In Y → X systems, a class-conditional shift occurs if the change in the joint probability 1 P(x∩y) is due to changes in the class-conditional distributions (right). The marginal probabilities of the labels (left) do not change. 19 [PITH_FULL_IMAGE:figures/full_fig_p01…
Figure 9
Figure 9. Figure 9: In a shared embedding approach, both the source and the target domain are transformed into a third domain that is considered [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]
Figure 10
Figure 10. Figure 10: A possible setup for domain adaptation is to simultaneously train the feature extractor and the classifier based on transformed source [PITH_FULL_IMAGE:figures/full_fig_p022_10.png]
Figure 11
Figure 11. Figure 11: In the technical risk & efficacy strategy proposed by Venable et al. (2016), early artificial testing guides the development process. For the overall evaluation of the framework, three episodes are conducted. The third EE (gray triangle) under real-life conditions was…
Figure 12
Figure 12. Figure 12: The overview of possible domain adaptation scenarios shows the occurring shifts for a simple one-dimensional example. Attributes that are given by the respective [PITH_FULL_IMAGE:figures/full_fig_p028_12.png]
Figure 13
Figure 13. Figure 13: For a Y → X scenario, a prior shift in the target domain results in a different concept being learned by a linear classifier. In this example the prior probabilities are Ps(y = +1) = Ps(y = −1) = 0.5 for the source domain and Pt(y = +1) = 0.75, Pt(y = −1) = 0.25 in th…
Figure 14
Figure 14. Figure 14: For a covariate shift, the correct concept can still be learned if enough sample data is available and if the model is correctly specified, [PITH_FULL_IMAGE:figures/full_fig_p030_14.png]
Figure 15
Figure 15. Figure 15: Benign case of a Y → X general shift. Although the source classifier is not optimal for the target domain, the resulting error is relatively small because (a) the classifier is robust for the class-conditional shift, and (b) fewer negative instances occur, which are r…
Figure 16
Figure 16. Figure 16: The performance of a misspecified classifier can be improved by sample-based reweighting of the source to resemble the target [PITH_FULL_IMAGE:figures/full_fig_p033_16.png]
Figure 17
Figure 17. Figure 17: Example of class ‘5’ from each data set. [PITH_FULL_IMAGE:figures/full_fig_p036_17.png]
Figure 18
Figure 18. Figure 18: The label distribution for MNIST and SVHN data set shows a class imbalance for SVHN. Because SVHN contains image data of [PITH_FULL_IMAGE:figures/full_fig_p037_18.png]
Figure 19
Figure 19. Figure 19: Between-subject evaluation study design. Two treatment groups with/ [PITH_FULL_IMAGE:figures/full_fig_p038_19.png]
Figure 20
Figure 20. Figure 20: Survey responses for technology acceptance model constructs. [PITH_FULL_IMAGE:figures/full_fig_p040_20.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 14 canonical work pages

  1. [6]

    The journal of machine learning research 17, 2096–2030

    Domain-adversarial training of neural networks. The journal of machine learning research 17, 2096–2030. Ghifary, M., Kleijn, W.B., Zhang, M., Balduzzi, D.,

    work page

  2. [8]

    Cycada: Cycle-consistent adversarial domain adaptation, in: International Conference on Machine Learning, PMLR. pp. 1989–1998. Hong, H.S., Kumar, A., Lee, D.G.,

    work page 1989

  3. [9]

    Augmented Cyclic Adversarial Learning for Low Resource Domain Adaptation

    Augmented cyclic adversarial learning for low resource domain adaptation. arXiv preprint arXiv:1807.00374 arXiv:1807.00374. Huang, J., Gretton, A., Borgwardt, K.M., Sch¨olkopf, B., Smola, A.J.,

    work page Pith review arXiv

  4. [10]

    Pearl, J.,

    1109/TKDE.2009.191. Pearl, J.,

    work page 2009

  5. [11]

    volume 6314, pp

    Springer Berlin Heidelberg, Berlin, Heidelberg. volume 6314, pp. 213–226. doi:10.1007/978-3-642-15561-1_16 . Iivari, J.,

    doi:10.1007/978-3-642-15561-1_16

  6. [13]

    Science 349, 255–260

    Machine learning: Trends, perspectives, and prospects. Science 349, 255–260. doi: 10.1126/science. aaa8415. Kamnitsas, K., Baumgartner, C., Ledig, C., Newcombe, V .F.J., Simpson, J.P., Kane, A.D., Menon, D.K., Nori, A., Criminisi, A., Rueckert, D., Glocker, B.,

    doi:10.1126/science

  7. [14]

    Unsupervised domain adaptation in brain lesion segmentation with adversarial networks

    Unsupervised domain adaptation in brain lesion segmentation with adversarial networks. arXiv:1612.08894 [cs] arXiv:1612.08894. Kang, G., Jiang, L., Yang, Y ., Hauptmann, A.G.,

    work page Pith review arXiv

  8. [16]

    arXiv:1812.11806 [cs, stat] arXiv:1812.11806

    An introduction to domain adaptation and transfer learning. arXiv:1812.11806 [cs, stat] arXiv:1812.11806. Kouw, W.M., Loog, M.,

    arXiv

Show all 37 references
  1. [18]

    arXiv:1811.05443 [cs, stat] arXiv:1811.05443

    Co-regularized Alignment for Unsupervised Domain Adaptation. arXiv:1811.05443 [cs, stat] arXiv:1811.05443. Lee, D., Seung, H.S.,

    arXiv

  2. [19]

    Image to Image Translation for Domain Adaptation, in: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, IEEE, Salt Lake City, UT. pp. 4500–4509. doi:10.1109/CVPR.2018.00473. Netzer, Y ., Wang, T., Coates, A., Bissacco, A., Wu, B., Ng, A.Y .,

    2018

  3. [20]

    IEEE Transactions on Medical Imaging 34, 1018–1030

    Transfer Learning Improves Supervised Image Segmentation Across Imaging Protocols. IEEE Transactions on Medical Imaging 34, 1018–1030. doi: 10.1109/TMI.2014.2366792. Palan, S., Schitter, C.,

    2014

  4. [23]

    2369–2372

    Adversarial Multi-Task Learning of Deep Neural Networks for Robust Speech Recognition, in: Interspeech 2016, pp. 2369–2372. doi: 10.21437/Interspeech.2016-879. Shu, R., Bui, H.H., Narui, H., Ermon, S.,

    2016 doi

  5. [24]

    arXiv:1802.08735 [cs, stat] arXiv:1802.08735

    A DIRT-T Approach to Unsupervised Domain Adaptation. arXiv:1802.08735 [cs, stat] arXiv:1802.08735. Silverman, B.W.,

    arXiv

  6. [25]

    British Journal of Management 16, 19–36

    Management Research as a Design Science: Articulating the Research Products of Mode 2 Knowledge Production in Management. British Journal of Management 16, 19–36. doi: 10.1111/j.1467-8551.2005.00437.x. Venable, J., Pries-Heje, J., Baskerville, R.,

    2005

  7. [26]

    European Journal of Informa- tion Systems 25, 77–89

    FEDS: A Framework for Evaluation in Design Science Research. European Journal of Informa- tion Systems 25, 77–89. doi: 10.1057/ejis.2014.36. Wang, M., Deng, W.,

    2014 doi

  8. [27]

    arXiv:1802.03601 [cs] arXiv:1802.03601

    Deep Visual Domain Adaptation: A Survey. arXiv:1802.03601 [cs] arXiv:1802.03601. Weiss, K., Khoshgoftaar, T.M., Wang, D.,

    arXiv

  9. [28]

    White, H.,

    doi:10.1186/s40537-016-0043-6 . White, H.,

    doi

  10. [31]

    Neural Computation 8, 1341–1390

    The Lack of A Priori Distinctions Between Learning Algorithms. Neural Computation 8, 1341–1390. doi:10.1162/neco. 1996.8.7.1341. Wu, Z., Zhang, H., Guo, J., Ji, Y ., Pecht, M.,

    1996 doi

  11. [33]

    arXiv:1903.04687 [cs] arXiv:1903.04687

    Transfer Adaptation Learning: A Decade Survey. arXiv:1903.04687 [cs] arXiv:1903.04687. Zhang, Y ., Barzilay, R., Jaakkola, T.,

    1903 arXiv

  12. [35]

    IEEE Access 6, 44698–44706

    Unsupervised Domain Adaptation by Mapped Correlation Alignment. IEEE Access 6, 44698–44706. doi:10.1109/ACCESS.2018.2865249. Zhao, H., Combes, R.T.D., Zhang, K., Gordon, G.,

    2018

  13. [36]

    arXiv:1911.02685 [cs, stat] arXiv:1911.02685

    A Comprehensive Survey on Transfer Learning. arXiv:1911.02685 [cs, stat] arXiv:1911.02685. Zliobaite, I.,

    1911 arXiv

  14. [37]

    CoRR abs /1010.4784

    Learning under concept drift: An overview. CoRR abs /1010.4784. arXiv:1010.4784. 48

    arXiv

  15. [46]

    Ganin, Y ., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., Marchand, M., Lempitsky, V .,

    doi: 10.1145/2523813. Ganin, Y ., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., Marchand, M., Lempitsky, V .,

    doi

  16. [188]

    Zhang, K., Sch ¨olkopf, B., Muandet, K., Wang, Z.,

    doi:10.1145/1291233.1291276. Zhang, K., Sch ¨olkopf, B., Muandet, K., Wang, Z.,

  17. [1994]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 16, 550–554

    A database for handwritten text recognition research. IEEE Transactions on Pattern Analysis and Machine Intelligence 16, 550–554. doi: 10.1109/34.291440. Hutchison, D., Kanade, T., Kittler, J., Kleinberg, J.M., Mattern, F.,

    doi

  18. [1995]

    Machine Learning 20, 273–297

    Support-vector networks. Machine Learning 20, 273–297. doi:10.1007/BF00994018. Elkan, C.,

    doi

  19. [1996]

    Machine Learning 23, 69–101

    Learning in the Presence of Concept Drift and Hidden Contexts. Machine Learning 23, 69–101. doi: 10.1023/A: 1018046501280. Wilson, G., Cook, D.J.,

    doi

  20. [2002]

    Neural Computation 14, 21–41

    Adjusting the Outputs of a Classifier to New aPriori Probabilities: A Simple Procedure. Neural Computation 14, 21–41. doi: 10.1162/089976602753284446. Schneider, S., Ecker, A.S., Macke, J.H., Bethge, M.,

    doi

  21. [2005]

    (Eds.), Information Systems Development

    Information Systems as a Design Science: Some concerns, in: Vasilecas, O., Wojtkowski, W., Zupan ˇciˇc, J., Caplinskas, A., Wo- jtkowski, W.G., Wrycza, S. (Eds.), Information Systems Development. Springer US, Boston, MA, pp. 15–27. doi:10.1007/0-387-28809-0_

    doi

  22. [2010]

    volume 22 of Integrated Series in Information Systems

    Design Research in Information Systems. volume 22 of Integrated Series in Information Systems. Springer US, Boston, MA. doi: 10.1007/978-1-4419-5653-8 . Hevner, A.R.,

    doi

  23. [2015]

    arXiv:1412.4446 [cs, stat] arXiv:1412.4446

    Domain-Adversarial Neural Networks. arXiv:1412.4446 [cs, stat] arXiv:1412.4446. Bashath, S., Perera, N., Tripathi, S., Manjang, K., Dehmer, M., Streib, F.E.,

    arXiv

  24. [2016]

    The Cityscapes Dataset for Semantic Urban Scene Understanding, in: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, Las Vegas, NV , USA. pp. 3213–3223. doi:10.1109/CVPR.2016.350. Cortes, C., Mansour, Y ., Mohri, M.,

    2016 doi

  25. [2017]

    Transactions of the Association for 47 Computational Linguistics 5, 515–528

    Aspect-augmented Adversarial Networks for Domain Adaptation. Transactions of the Association for 47 Computational Linguistics 5, 515–528. doi: 10.1162/tacl_a_00077. Zhang, Y ., Wang, N., Cai, S., Song, L.,

    doi

  26. [2018]

    arXiv:1706.05208 [cs] arXiv:1706.05208

    Self-ensembling for visual domain adaptation. arXiv:1706.05208 [cs] arXiv:1706.05208. Gama, J., ˇZliobaitundefined, I., Bifet, A., Pechenizkiy, M., Bouchachia, A.,

    arXiv

  27. [2019]

    Contrastive Adaptation Network for Unsupervised Domain Adaptation, in: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, Long Beach, CA, USA. pp. 4888–4897. doi: 10.1109/ CVPR.2019.00503. Karimian, M., Beigy, H.,

    2019

  28. [2020]

    arXiv:1812.02849 [cs, stat] arXiv:1812.02849

    A Survey of Unsupervised Deep Domain Adaptation. arXiv:1812.02849 [cs, stat] arXiv:1812.02849. Wolberg, W.H., Street, W.N., Mangasarian, O.L.,

    arXiv

  29. [2021]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 43, 766–785

    A review of domain adaptation without target labels. IEEE Transactions on Pattern Analysis and Machine Intelligence 43, 766–785. doi: 10.1109/TPAMI.2019.2945942. 45 Kumar, A., Sattigeri, P., Wadhawan, K., Karlinsky, L., Feris, R., Freeman, W.T., Wornell, G.,

    2019

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.