phenomenon · claim-type · cross-domain-transfer · mechanism · claim-relation · replication-status · limitation — grounded in the full text| phenomenon | improving a target learner using information from a different but related source domain— “Therefore, there is a need to create high-performance learners trained with more easily obtained data from different domains. This methodology is referred to as transfer learning.” | |
| claim-type | critique— “Throughout the literature on transfer learning, there are a number of terminology inconsistencies. Phrases such as transfer learning and domain adaptation are used to refer to similar processes. The following definitions will be used in this paper.” | |
| method — evidence | — “The majority of the homogeneous transfer learning solutions employ one of three general strategies which include trying to correct for the marginal distribution difference in the source, trying to correct for the conditional distribution difference in the source, or trying to correct both the marginal and conditional distribution differences… | |
| object of study — evidence | — “This survey paper formally defines transfer learning, presents information on current solutions, and reviews applications applied to transfer learning.” | |
| research question — evidence | — “An assumption of traditional machine learning methodologies is the training data and testing data are taken from the same domain, such that the input feature space and data distribution characteristics are the same.” | |
| contribution — evidence | — “This survey paper formally defines transfer learning, presents information on current solutions, and reviews applications applied to transfer learning.” | |
| mechanism | WHY TRANSFER LEARNING EXISTS AT ALL: ordinary machine learning assumes train and test come from one domain with one distribution, and the method is what you do when that assumption is false and the target data is expensive to collect— “An assumption of traditional machine learning methodologies is the training data and testing data are taken from the same domain, such that the input feature space and data distribution characteristics are the same.” | mechanisms[0] |
| mechanism | NEGATIVE TRANSFER IS DEFINED BY COMPARISON, NOT BY DAMAGE: a learner trained on the target alone is compared against one trained with the source added, and transfer is negative when the plain learner wins— “negative transfer occurs when the performance of fT1(·) is greater than the performance of fT2(·). The topic of negative transfer addresses the need to quantify the amount of relatedness between the source domain and the target domain and whether an attempt to transfer knowledge from the source domain should be made.” | mechanisms[1] |
| mechanism | PSEUDO LABELS CORRECT A DISTRIBUTION USING GUESSES AT THE THING BEING CORRECTED FOR: with no labeled target data, labels are estimated and then used to fix the conditional distribution - which the paper itself calls problematic— “To address the lack of labeled target data, some solutions estimate the labels for the target data (called pseudo labels), which are then used to correct the conditional distribution differences. This method is problematic because the conditional distribution corrections are being made with the aid of pseudo labels.” | mechanisms[2] |
| mechanism | instance transfer reweights source examples to correct MARGINAL distribution differences, and works best precisely when the CONDITIONAL distribution is already the same— “A common method used in this case is for instances from the source domain to be reweighted in an attempt to correct for marginal distribution differences.” | mechanisms[3] |
| mechanism | the two feature approaches differ in direction: one bends the source onto the target, the other maps both into a shared latent space— “Feature-based transfer learning approaches are categorized in two ways.” | mechanisms[4] |
| mechanism | parameter transfer shares model parameters or combines multiple reweighted source learners into an ensemble for the target— “The third transfer category is to transfer knowledge through shared parameters of source and target domain learner models or by creating multiple source learner models and optimally combining the reweighted learners (ensemble learners) to form an improved target learner (examples in Gao [37], Bonilla [8], and Evgeniou [33]).” | mechanisms[5] |
| mechanism | source classifiers are weighted by how close their conditional distribution is to the target, summed, and used to manufacture pseudo labels for unlabelled target data— “This is accomplished by first building a classifier for each separate source domain.” | mechanisms[6] |
| mechanism | THE FIELD IS DECLARED TERMINOLOGICALLY INCONSISTENT AND THE PAPER SAYS SO IN ITS OWN VOICE: `Throughout the literature on transfer learning, there are a number of terminology inconsistencies. Phrases such as transfer learning and domain adaptation are used to refer to similar processes. The following definitions will be used in this paper.` A survey opening by stipulating the vocabulary because the literature has none.— “Throughout the literature on transfer learning, there are a number of terminology inconsistencies. Phrases such as transfer learning and domain adaptation are used to refer to similar processes. The following definitions will be used in this paper” | mechanisms[7] |
| mechanism | THE SAME CONFIGURATION, RENAMED IN THE NEXT SENTENCE: `In Gong [42] and Blitzer [5], semi-supervised transfer learning is the case of having abundant labeled source data and limited labeled target data.` The sentence immediately before it - quoted separately on this record - gives that exact configuration to Daume and Chattopadhyay as SUPERVISED. The full four-scheme enumeration spans several sentences with a journal running header splicing one of them, so it is carried across four tags on this record rather than asserted from one span.— “In Gong [42] and Blitzer [5], semi-supervised transfer learning is the case of having abundant labeled source data and limited labeled target data” | mechanisms[8] |
| mechanism | AND THE PAPER'S REPAIR IS TO REFUSE THE VOCABULARY ALTOGETHER: `This paper will explicitly state when labeled and unlabeled data are being used in the source and target domains.` It does not pick a winner and does not coin a fifth term - it describes the configuration and drops the names. THE SAME MOVE CONSORT MADE IN 2010, replacing `intention to treat` with `a more explicit request for information` (schulz2010).— “This paper will explicitly state when labeled and unlabeled data are being used in the source and target domains” | mechanisms[9] |
| sense | supervised transfer learning - DAUME AND CHATTOPADHYAY'S SENSE: abundant labeled source data and LIMITED labeled target data— “For example, Daumé [22] and Chattopadhyay [14] define supervised transfer learning as the case of having abundant labeled source data and limited labeled target data, and semi-supervised transfer learning as the case of abundant labeled source data and no labeled target data.” | senses[0] |
| sense | semi-supervised transfer learning - DAUME AND CHATTOPADHYAY'S SENSE: abundant labeled source and NO labeled target. Gong and Blitzer use this same phrase for what Daume calls SUPERVISED - the two labels are swapped between literatures— “For example, Daumé [22] and Chattopadhyay [14] define supervised transfer learning as the case of having abundant labeled source data and limited labeled target data, and semi-supervised transfer learning as the case of abundant labeled source data and no labeled target data.” | senses[1] |
| sense | unsupervised transfer learning - PAN'S SENSE: NO labeled source and NO labeled target, where the same phrase elsewhere in the literature means abundant labeled source and no labeled target. THREE GROUPS, ONE VOCABULARY, THREE MAPPINGS— “Pan [84] refers to inductive transfer learning as the case of having available labeled target domain data, transductive transfer learning as the case of having labeled source and no labeled target domain data, and unsupervised transfer learning as the case of having no labeled source and no labeled target domain data.” | senses[2] |
| sense | transfer learning - the paper's response to its own finding: it refuses the contested labels entirely and states the data conditions each time instead of naming them— “This paper will explicitly state when labeled and unlabeled data are being used in the source and target domains.” | senses[3] |
| sense | `supervised transfer learning` - DAUME AND CHATTOPADHYAY'S SENSE: `abundant labeled source data and LIMITED labeled target data`. Gong and Blitzer give that same configuration to SEMI-SUPERVISED, and Cook and Feuz call it INFORMED SUPERVISED.— “For example, Daumé [22] and Chattopadhyay [14] define supervised transfer learning as the case of having abundant labeled source data and limited labeled target data, and semi-supervised transfer learning as the case of abundant labeled source data and no labeled target data” | senses[4] |
| sense | `semi-supervised transfer learning` - GONG AND BLITZER'S SENSE: `abundant labeled source data and LIMITED labeled target data`. THAT IS THE CONFIGURATION Daume and Chattopadhyay CALL SUPERVISED - see senses[4] and the `daume` mechanism on this record, which quote their sentence. One name over two configurations and one configuration under two names, in consecutive sentences.— “In Gong [42] and Blitzer [5], semi-supervised transfer learning is the case of having abundant labeled source data and limited labeled target data” | senses[5] |
| sense | `supervised` / `unsupervised` - COOK AND FEUZ MOVE THE AXIS: the pair refers to `the presence or absence of labeled data in the source domain` and a SECOND pair, `informed` / `uninformed`, carries the target. Two orthogonal axes where the other schemes have one.— “Cook [19] and Feuz [36] provide a different variation where the definition of supervised or unsupervised refers to the presence or absence of labeled data in the source domain and informed or uninformed refers to the presence or absence of labeled data in the target domain. With this definition, a labeled source and limited labeled target… | senses[6] |
| sense | `domain adaptation` - AS STIPULATED HERE, because the field does not agree: `the process of adapting one or more source domains for the means of transferring information to improve the performance of a target learner`, by bringing the source's distribution closer to the target's.— “Domain adaptation, as it pertains to transfer learning, is the process of adapting one or more source domains for the means of transferring information to improve the performance of a target learner. The domain adaptation process attempts to alter a source domain in an attempt to bring the distribution of the source closer to that of the… | senses[7] |
| weight | domain adaptation —part-of→ transfer learning — 0.9— “Domain adaptation, as it pertains to transfer learning, is the process of adapting one or more source domains for the means of transferring information to improve the performance of a target learner.” | concept_edges[0] |
| audit | negative transfer —distinct-from→ positive transfer — Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but endpoint by anaphora: 'the definition above' = negative transfer (previous sentences); distinct-from is loose.— “Extending the definition above, positive transfer occurs when the performance of fT2(·) is greater than the performance of fT1(·).” | concept_edges[1] |
| edge | feature space —part-of→ domain D— “A domain D is defined by two parts, a feature space X and a marginal probability distribution P(X)” | concept_edges[2] |
| edge | label space —part-of→ task T— “For a given domain D, a task T is defined by two parts, a label space Y, and a predictive function f(·), which is learned from the feature vector and label pairs” | concept_edges[3] |
| edge | predictive function —part-of→ task T— “For a given domain D, a task T is defined by two parts, a label space Y, and a predictive function f(·), which is learned from the feature vector and label pairs” | concept_edges[4] |
| limitation | THE LITERATURE'S VOCABULARY IS INCONSISTENT AND THE PAPER SAYS SO: transfer learning and domain adaptation are used for similar processes, and the supervised / semi-supervised / unsupervised labels mean different things to different authors— “Throughout the literature on transfer learning, there are a number of terminology inconsistencies. Phrases such as transfer learning and domain adaptation are used to refer to similar processes. The following definitions will be used in this paper.” | limitations[0] |
| limitation | Some surveyed solutions are tied to their application and are stated not to generalise to others— “In some cases, the proposed transfer learning solutions are very specific to the underlying application and cannot be generically used for other applications.” | limitations[1] |
| limitation | Correcting the conditional distribution rests on pseudo labels, which the paper names as problematic because the correction uses estimates of the very labels it lacks— “To address the lack of labeled target data, some solutions estimate the labels for the target data (called pseudo labels), which are then used to correct the conditional distribution differences. This method is problematic because the conditional distribution corrections are being made with the aid of pseudo labels.” | limitations[2] |
| limitation | No single repository exists for the surveyed solutions, so the software behind the literature is scattered - the paper asks for one— “A single open-source software repository for published transfer learning solutions would be a great asset to the research community.” | limitations[3] |
| limitation | heterogeneous solutions mostly assume the domain distributions are the SAME, and need further adaptation steps when they are not— “The majority of the heterogeneous transfer learning solutions are focused on aligning the input spaces of the source and target domains with the assumption that the domain distributions are the same.” | limitations[4] |
| limitation | transferring on a defined relationship between domains is named as the least used of the four categories— “The last transfer category (and the least used approach) is to transfer knowledge based on some defined relationship between the source and target domains (examples in Mihalkova [74] and Li [62]).” | limitations[5] |
finding · quantity · variable · definition · key-term — grounded in the full text| finding | in the muscle-fatigue test EVERY transfer learning approach beat the non-transfer baseline, with the two proposed methods ranking first and second— “The order of performance from best to worst is 2SW-MDA, CP-MDA, Duan [29], Zhong [143], Gao [37], Pan [87], Huang [51], and the baseline approach.” | content_tags.findings[0] |
| finding | an often-cited domain adaptation method is reported as requiring only ten lines of Perl - the simplicity is the claim— “In an early and often cited work, Daumé [22] proposes a simple domain adaptation algorithm, referred to as the feature augmentation method (FAM), requiring only ten lines of Perl script that uses labeled source data and limited labeled target data.” | content_tags.findings[1] |
| finding | Homogeneous transfer learning is directly applicable to big data environments.— “The methodology of homogeneous transfer learning is directly applicable to a big data environment.” | content_tags.findings[2] |
| definition | transfer categories— “The form of information transfer is categorized into four general Transfer Categories [84].” | content_tags.definitions[0] |
| definition | context feature bias— “The issue is referred to as context feature bias, which causes the conditional distributions between the source and target domains to be different.” | content_tags.definitions[1] |
| concept | transfer learning— “Therefore, there is a need to create high-performance learners trained with more easily obtained data from different domains. This methodology is referred to as transfer learning.” | concepts[0] |
| concept | domain adaptation— “Domain adaptation, as it pertains to transfer learning, is the process of adapting one or more source domains for the means of transferring information to improve the performance of a target learner.” | concepts[1] |
| concept | negative transfer— “negative transfer occurs when the performance of fT1(·) is greater than the performance of fT2(·). The topic of negative transfer addresses the need to quantify the amount of relatedness between the source domain and the target domain and whether an attempt to transfer knowledge from the source domain should be made.” | concepts[2] |
| concept | positive transfer— “Extending the definition above, positive transfer occurs when the performance of fT2(·) is greater than the performance of fT1(·).” | concepts[3] |
| concept | feature space— “A domain D is defined by two parts, a feature space X and a marginal probability distribution P(X)” | concepts[4] |
| concept | domain D— “A domain D is defined by two parts, a feature space X and a marginal probability distribution P(X)” | concepts[5] |
| concept | task T— “For a given domain D, a task T is defined by two parts, a label space Y, and a predictive function f(·), which is learned from the feature vector and label pairs” | concepts[6] |
| concept | label space— “For a given domain D, a task T is defined by two parts, a label space Y, and a predictive function f(·), which is learned from the feature vector and label pairs” | concepts[7] |
| concept | predictive function— “For a given domain D, a task T is defined by two parts, a label space Y, and a predictive function f(·), which is learned from the feature vector and label pairs” | concepts[8] |
| concept | supervised transfer learning— “For example, Daumé [22] and Chattopadhyay [14] define supervised transfer learning as the case of having abundant labeled source data and limited labeled target data, and semi-supervised transfer learning as the case of abundant labeled source data and no labeled target data.” | concepts[9] |
| concept | semi-supervised transfer learning— “For example, Daumé [22] and Chattopadhyay [14] define supervised transfer learning as the case of having abundant labeled source data and limited labeled target data, and semi-supervised transfer learning as the case of abundant labeled source data and no labeled target data.” | concepts[10] |
| concept | inductive transfer learning— “Pan [84] refers to inductive transfer learning as the case of having available labeled target domain data, transductive transfer learning as the case of having labeled source and no labeled target domain data, and unsupervised transfer learning as the case of having no labeled source and no labeled target domain data.” | concepts[11] |
| concept | transductive transfer learning— “Pan [84] refers to inductive transfer learning as the case of having available labeled target domain data, transductive transfer learning as the case of having labeled source and no labeled target domain data, and unsupervised transfer learning as the case of having no labeled source and no labeled target domain data.” | concepts[12] |
| concept | unsupervised transfer learning— “Pan [84] refers to inductive transfer learning as the case of having available labeled target domain data, transductive transfer learning as the case of having labeled source and no labeled target domain data, and unsupervised transfer learning as the case of having no labeled source and no labeled target domain data.” | concepts[13] |
| concept | terminology inconsistencies— “Throughout the literature on transfer learning, there are a number of terminology inconsistencies. Phrases such as transfer learning and domain adaptation are used to refer to similar processes. The following definitions will be used in this paper.” | concepts[14] |
| concept | pseudo labels— “To address the lack of labeled target data, some solutions estimate the labels for the target data (called pseudo labels), which are then used to correct the conditional distribution differences. This method is problematic because the conditional distribution corrections are being made with the aid… | concepts[15] |
| concept | marginal distribution | concepts[16] |
| concept | conditional distribution | concepts[17] |
| concept | source domain | concepts[18] |
| concept | target domain | concepts[19] |
| concept | context feature bias — not defined in this paper's own words; THE SAME FEATURE MEANING SOMETHING DIFFERENT IN THE SOURCE AND TARGET DOMAIN - polysemy inside a feature space, and the reason the conditional distributions diverge Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[20] |
| concept | Transfer Categories — not defined in this paper's own words; the four-way classification of WHAT is transferred - instances, features, parameters, or a defined relation between domains Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[21] |
| concept | asymmetric feature transformation — not defined in this paper's own words; reweighting the SOURCE features to look like the target - the source is bent toward the target Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[22] |
| concept | symmetric feature transformation — not defined in this paper's own words; mapping BOTH domains into a common latent feature space instead of bending one onto the other Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[23] |
| concept | feature augmentation method — not defined in this paper's own words; resolving context feature bias by triplicating the feature set into common, source-specific and target-specific copies Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[24] |
| concept | informed supervised transfer learning — not defined in this paper's own words; Cook and Feuz's name for the configuration Daume calls SUPERVISED and Gong calls SEMI-SUPERVISED: abundant labelled source, limited labelled target. Three names, one setup. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[25] |
| concept | transductive transfer learning — not defined in this paper's own words; Pan's name for labelled source and no labelled target - the configuration Daume calls SEMI-SUPERVISED and Gong calls UNSUPERVISED Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[26] |
When we have changed our mind. Some entries in the table begin with a word in capitals and the reason for it. EVIDENCE CORRECTED — the finding still stands, but we had quoted the wrong sentence from the paper, and the quotation has been replaced with the right one. RETRACTED — the entry itself was wrong; it stays on the page, marked, with the reason, rather than being deleted. WITHDRAWN AND REPLACED — the entry was wrong, and a corrected version of it appears in the entries below. Each mark carries its own reason, in brackets, because the reasons differ from one entry to the next. An entry we have WITHDRAWN is removed from this page and kept in the record, with its reason. A line reading Audit and a date is a later re-reading of the entry against its whole sentence: DOUBTFUL means it was kept, and the reason says what is loose about it. Quotation marks inside the tagger's sentence. Where a tag is the tagger's own sentence, a phrase in quotation marks inside it is borrowed from the paper — the tagger's wording around it is not. The paper's own sentence is the evidence quote for that row; on this page it is withheld with the article, and the article is at its publisher. Nothing inside quotation marks anywhere on this page is ours: if it is quoted, it is the paper's. Whose words a concept carries. A concept entry records a term the paper itself uses, so the term, the sense given for it, and its evidence are the paper's own words, quoted where the paper fixes them. This is the opposite of a sense entry in the tag table, which is the tagger's sentence about how the paper uses a term. Where a concept entry could not be grounded in the paper's words, the entry says so on its own row.
Tag layer © the Meta-Tagging Project, CC BY 4.0. The paper text is © its authors/publisher
and is linked, not reproduced. Excerpts shown are short quotations for identification and verification.
· back to index
· concepts · authors
· עברית
© 2026 Shir Sivroni · ORCID 0009-0000-2597-2824