Minimum Class Confusion for Versatile Domain Adaptation View Full Text


Ontology type: schema:Chapter      Open Access: True


Chapter Info

DATE

2020-11-12

AUTHORS

Ying Jin , Ximei Wang , Mingsheng Long , Jianmin Wang

ABSTRACT

There are a variety of Domain Adaptation (DA) scenarios subject to label sets and domain configurations, including closed-set and partial-set DA, as well as multi-source and multi-target DA. It is notable that existing DA methods are generally designed only for a specific scenario, and may underperform for scenarios they are not tailored to. To this end, this paper studies Versatile Domain Adaptation (VDA), where one method can handle several different DA scenarios without any modification. Towards this goal, a more general inductive bias other than the domain alignment should be explored. We delve into a missing piece of existing methods: class confusion, the tendency that a classifier confuses the predictions between the correct and ambiguous classes for target examples, which is common in different DA scenarios. We uncover that reducing such pairwise class confusion leads to significant transfer gains. With this insight, we propose a general loss function: Minimum Class Confusion (MCC). It can be characterized as (1) a non-adversarial DA method without explicitly deploying domain alignment, enjoying faster convergence speed; (2) a versatile approach that can handle four existing scenarios: Closed-Set, Partial-Set, Multi-Source, and Multi-Target DA, outperforming the state-of-the-art methods in these scenarios, especially on one of the largest and hardest datasets to date (7.3%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$7.3\%$$\end{document} on DomainNet). Its versatility is further justified by two scenarios proposed in this paper: Multi-Source Partial DA and Multi-Target Partial DA. In addition, it can also be used as a general regularizer that is orthogonal and complementary to a variety of existing DA methods, accelerating convergence and pushing these readily competitive methods to stronger ones. Code is available at https://github.com/thuml/Versatile-Domain-Adaptation. More... »

PAGES

464-480

Identifiers

URI

http://scigraph.springernature.com/pub.10.1007/978-3-030-58589-1_28

DOI

http://dx.doi.org/10.1007/978-3-030-58589-1_28

DIMENSIONS

https://app.dimensions.ai/details/publication/pub.1132537449


Indexing Status Check whether this publication has been indexed by Scopus and Web Of Science using the SN Indexing Status Tool
Incoming Citations Browse incoming citations for this publication using opencitations.net

JSON-LD is the canonical representation for SciGraph data.

TIP: You can open this SciGraph record using an external JSON-LD service: JSON-LD Playground Google SDTT

[
  {
    "@context": "https://springernature.github.io/scigraph/jsonld/sgcontext.json", 
    "about": [
      {
        "id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/08", 
        "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
        "name": "Information and Computing Sciences", 
        "type": "DefinedTerm"
      }, 
      {
        "id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/0801", 
        "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
        "name": "Artificial Intelligence and Image Processing", 
        "type": "DefinedTerm"
      }
    ], 
    "author": [
      {
        "affiliation": {
          "alternateName": "School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China", 
          "id": "http://www.grid.ac/institutes/grid.12527.33", 
          "name": [
            "School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Jin", 
        "givenName": "Ying", 
        "id": "sg:person.07422717707.09", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.07422717707.09"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China", 
          "id": "http://www.grid.ac/institutes/grid.12527.33", 
          "name": [
            "School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Wang", 
        "givenName": "Ximei", 
        "id": "sg:person.014716677221.46", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.014716677221.46"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China", 
          "id": "http://www.grid.ac/institutes/grid.12527.33", 
          "name": [
            "School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Long", 
        "givenName": "Mingsheng", 
        "id": "sg:person.013417115303.81", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013417115303.81"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China", 
          "id": "http://www.grid.ac/institutes/grid.12527.33", 
          "name": [
            "School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Wang", 
        "givenName": "Jianmin", 
        "id": "sg:person.012303351315.43", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012303351315.43"
        ], 
        "type": "Person"
      }
    ], 
    "datePublished": "2020-11-12", 
    "datePublishedReg": "2020-11-12", 
    "description": "There are a variety of Domain Adaptation (DA) scenarios subject to label sets and domain configurations, including closed-set and partial-set DA, as well as multi-source and multi-target DA. It is notable that existing DA methods are generally designed only for a specific scenario, and may underperform for scenarios they are not tailored to. To this end, this paper studies Versatile Domain Adaptation (VDA), where one method can handle several different DA scenarios without any modification. Towards this goal, a more general inductive bias other than the domain alignment should be explored. We delve into a missing piece of existing methods: class confusion, the tendency that a classifier confuses the predictions between the correct and ambiguous classes for target examples, which is common in different DA scenarios. We uncover that reducing such pairwise class confusion leads to significant transfer gains. With this insight, we propose a general loss function: Minimum Class Confusion (MCC). It can be characterized as (1) a non-adversarial DA method without explicitly deploying domain alignment, enjoying faster convergence speed; (2) a versatile approach that can handle four existing scenarios: Closed-Set, Partial-Set, Multi-Source, and Multi-Target DA, outperforming the state-of-the-art methods in these scenarios, especially on one of the largest and hardest datasets to date (7.3%\\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym}\n\t\t\t\t\\usepackage{amsfonts}\n\t\t\t\t\\usepackage{amssymb}\n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$7.3\\%$$\\end{document} on DomainNet). Its versatility is further justified by two scenarios proposed in this paper: Multi-Source Partial DA and Multi-Target Partial DA. In addition, it can also be used as a general regularizer that is orthogonal and complementary to a variety of existing DA methods, accelerating convergence and pushing these readily competitive methods to stronger ones. Code is available at https://github.com/thuml/Versatile-Domain-Adaptation.", 
    "editor": [
      {
        "familyName": "Vedaldi", 
        "givenName": "Andrea", 
        "type": "Person"
      }, 
      {
        "familyName": "Bischof", 
        "givenName": "Horst", 
        "type": "Person"
      }, 
      {
        "familyName": "Brox", 
        "givenName": "Thomas", 
        "type": "Person"
      }, 
      {
        "familyName": "Frahm", 
        "givenName": "Jan-Michael", 
        "type": "Person"
      }
    ], 
    "genre": "chapter", 
    "id": "sg:pub.10.1007/978-3-030-58589-1_28", 
    "inLanguage": "en", 
    "isAccessibleForFree": true, 
    "isPartOf": {
      "isbn": [
        "978-3-030-58588-4", 
        "978-3-030-58589-1"
      ], 
      "name": "Computer Vision \u2013 ECCV 2020", 
      "type": "Book"
    }, 
    "keywords": [
      "domain adaptation", 
      "domain adaptation scenarios", 
      "label set", 
      "DA methods", 
      "DA scenarios", 
      "inductive bias", 
      "domain alignment", 
      "class confusion", 
      "faster convergence speed", 
      "art methods", 
      "hard datasets", 
      "adaptation scenarios", 
      "specific scenarios", 
      "target examples", 
      "loss function", 
      "convergence speed", 
      "Multi-Source", 
      "competitive methods", 
      "scenarios", 
      "ambiguous classes", 
      "general loss functions", 
      "partial set", 
      "set", 
      "classifier", 
      "dataset", 
      "regularizer", 
      "method", 
      "alignment", 
      "code", 
      "variety", 
      "adaptation", 
      "goal", 
      "example", 
      "speed", 
      "convergence", 
      "configuration", 
      "end", 
      "pieces", 
      "confusion", 
      "prediction", 
      "class", 
      "versatility", 
      "strong ones", 
      "one", 
      "gain", 
      "insights", 
      "function", 
      "versatile approach", 
      "state", 
      "DA", 
      "modification", 
      "date", 
      "addition", 
      "bias", 
      "tendency", 
      "transfer gain", 
      "approach", 
      "paper"
    ], 
    "name": "Minimum Class Confusion for Versatile Domain Adaptation", 
    "pagination": "464-480", 
    "productId": [
      {
        "name": "dimensions_id", 
        "type": "PropertyValue", 
        "value": [
          "pub.1132537449"
        ]
      }, 
      {
        "name": "doi", 
        "type": "PropertyValue", 
        "value": [
          "10.1007/978-3-030-58589-1_28"
        ]
      }
    ], 
    "publisher": {
      "name": "Springer Nature", 
      "type": "Organisation"
    }, 
    "sameAs": [
      "https://doi.org/10.1007/978-3-030-58589-1_28", 
      "https://app.dimensions.ai/details/publication/pub.1132537449"
    ], 
    "sdDataset": "chapters", 
    "sdDatePublished": "2022-05-20T07:46", 
    "sdLicense": "https://scigraph.springernature.com/explorer/license/", 
    "sdPublisher": {
      "name": "Springer Nature - SN SciGraph project", 
      "type": "Organization"
    }, 
    "sdSource": "s3://com-springernature-scigraph/baseset/20220519/entities/gbq_results/chapter/chapter_342.jsonl", 
    "type": "Chapter", 
    "url": "https://doi.org/10.1007/978-3-030-58589-1_28"
  }
]
 

Download the RDF metadata as:  json-ld nt turtle xml License info

HOW TO GET THIS DATA PROGRAMMATICALLY:

JSON-LD is a popular format for linked data which is fully compatible with JSON.

curl -H 'Accept: application/ld+json' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-58589-1_28'

N-Triples is a line-based linked data format ideal for batch operations.

curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-58589-1_28'

Turtle is a human-readable linked data format.

curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-58589-1_28'

RDF/XML is a standard XML format for linked data.

curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-58589-1_28'


 

This table displays all metadata directly associated to this object as RDF triples.

154 TRIPLES      23 PREDICATES      83 URIs      76 LITERALS      7 BLANK NODES

Subject Predicate Object
1 sg:pub.10.1007/978-3-030-58589-1_28 schema:about anzsrc-for:08
2 anzsrc-for:0801
3 schema:author Nf9d6900b23f14f5990142929f38ba21d
4 schema:datePublished 2020-11-12
5 schema:datePublishedReg 2020-11-12
6 schema:description There are a variety of Domain Adaptation (DA) scenarios subject to label sets and domain configurations, including closed-set and partial-set DA, as well as multi-source and multi-target DA. It is notable that existing DA methods are generally designed only for a specific scenario, and may underperform for scenarios they are not tailored to. To this end, this paper studies Versatile Domain Adaptation (VDA), where one method can handle several different DA scenarios without any modification. Towards this goal, a more general inductive bias other than the domain alignment should be explored. We delve into a missing piece of existing methods: class confusion, the tendency that a classifier confuses the predictions between the correct and ambiguous classes for target examples, which is common in different DA scenarios. We uncover that reducing such pairwise class confusion leads to significant transfer gains. With this insight, we propose a general loss function: Minimum Class Confusion (MCC). It can be characterized as (1) a non-adversarial DA method without explicitly deploying domain alignment, enjoying faster convergence speed; (2) a versatile approach that can handle four existing scenarios: Closed-Set, Partial-Set, Multi-Source, and Multi-Target DA, outperforming the state-of-the-art methods in these scenarios, especially on one of the largest and hardest datasets to date (7.3%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$7.3\%$$\end{document} on DomainNet). Its versatility is further justified by two scenarios proposed in this paper: Multi-Source Partial DA and Multi-Target Partial DA. In addition, it can also be used as a general regularizer that is orthogonal and complementary to a variety of existing DA methods, accelerating convergence and pushing these readily competitive methods to stronger ones. Code is available at https://github.com/thuml/Versatile-Domain-Adaptation.
7 schema:editor N4dbadb9b1c6d4d829ce60368b1245bb3
8 schema:genre chapter
9 schema:inLanguage en
10 schema:isAccessibleForFree true
11 schema:isPartOf Nbf23f665a4444f1b963cbc72a2b3bad6
12 schema:keywords DA
13 DA methods
14 DA scenarios
15 Multi-Source
16 adaptation
17 adaptation scenarios
18 addition
19 alignment
20 ambiguous classes
21 approach
22 art methods
23 bias
24 class
25 class confusion
26 classifier
27 code
28 competitive methods
29 configuration
30 confusion
31 convergence
32 convergence speed
33 dataset
34 date
35 domain adaptation
36 domain adaptation scenarios
37 domain alignment
38 end
39 example
40 faster convergence speed
41 function
42 gain
43 general loss functions
44 goal
45 hard datasets
46 inductive bias
47 insights
48 label set
49 loss function
50 method
51 modification
52 one
53 paper
54 partial set
55 pieces
56 prediction
57 regularizer
58 scenarios
59 set
60 specific scenarios
61 speed
62 state
63 strong ones
64 target examples
65 tendency
66 transfer gain
67 variety
68 versatile approach
69 versatility
70 schema:name Minimum Class Confusion for Versatile Domain Adaptation
71 schema:pagination 464-480
72 schema:productId Na73d0bed8cdf45cfb47b7aa8b35288e6
73 Nc160c42da7c24f76a6b5e5544e8f48c7
74 schema:publisher N2dc07d51c5cd423b946fbfec859f83b5
75 schema:sameAs https://app.dimensions.ai/details/publication/pub.1132537449
76 https://doi.org/10.1007/978-3-030-58589-1_28
77 schema:sdDatePublished 2022-05-20T07:46
78 schema:sdLicense https://scigraph.springernature.com/explorer/license/
79 schema:sdPublisher Nbfa47e18ab854f8ba27ebf7667fb8df4
80 schema:url https://doi.org/10.1007/978-3-030-58589-1_28
81 sgo:license sg:explorer/license/
82 sgo:sdDataset chapters
83 rdf:type schema:Chapter
84 N2a239c02b6be4c938d65e4df53294d11 rdf:first sg:person.014716677221.46
85 rdf:rest Ne60b8159b4084ff88635d3644e5293bb
86 N2dc07d51c5cd423b946fbfec859f83b5 schema:name Springer Nature
87 rdf:type schema:Organisation
88 N4dbadb9b1c6d4d829ce60368b1245bb3 rdf:first Naf07f401f01d4c7fbe2e0e76e6ce0f55
89 rdf:rest Na606f05398574830b22ae08743f45bf8
90 N68b87a4dbd3045f1b2ee5fa683540c8e schema:familyName Bischof
91 schema:givenName Horst
92 rdf:type schema:Person
93 N6c74bc72ff4047439791c21d9f59de6d rdf:first Nd2a7d71235e34070bf7851a0de5ab33f
94 rdf:rest rdf:nil
95 N805d0a95748e473eac99450247f3ab7c schema:familyName Brox
96 schema:givenName Thomas
97 rdf:type schema:Person
98 Na606f05398574830b22ae08743f45bf8 rdf:first N68b87a4dbd3045f1b2ee5fa683540c8e
99 rdf:rest Nbe30529795e74ecf9ace13cb47a3c997
100 Na73d0bed8cdf45cfb47b7aa8b35288e6 schema:name doi
101 schema:value 10.1007/978-3-030-58589-1_28
102 rdf:type schema:PropertyValue
103 Nad85bdd4567c46d8b3880d5d43d0498b rdf:first sg:person.012303351315.43
104 rdf:rest rdf:nil
105 Naf07f401f01d4c7fbe2e0e76e6ce0f55 schema:familyName Vedaldi
106 schema:givenName Andrea
107 rdf:type schema:Person
108 Nbe30529795e74ecf9ace13cb47a3c997 rdf:first N805d0a95748e473eac99450247f3ab7c
109 rdf:rest N6c74bc72ff4047439791c21d9f59de6d
110 Nbf23f665a4444f1b963cbc72a2b3bad6 schema:isbn 978-3-030-58588-4
111 978-3-030-58589-1
112 schema:name Computer Vision – ECCV 2020
113 rdf:type schema:Book
114 Nbfa47e18ab854f8ba27ebf7667fb8df4 schema:name Springer Nature - SN SciGraph project
115 rdf:type schema:Organization
116 Nc160c42da7c24f76a6b5e5544e8f48c7 schema:name dimensions_id
117 schema:value pub.1132537449
118 rdf:type schema:PropertyValue
119 Nd2a7d71235e34070bf7851a0de5ab33f schema:familyName Frahm
120 schema:givenName Jan-Michael
121 rdf:type schema:Person
122 Ne60b8159b4084ff88635d3644e5293bb rdf:first sg:person.013417115303.81
123 rdf:rest Nad85bdd4567c46d8b3880d5d43d0498b
124 Nf9d6900b23f14f5990142929f38ba21d rdf:first sg:person.07422717707.09
125 rdf:rest N2a239c02b6be4c938d65e4df53294d11
126 anzsrc-for:08 schema:inDefinedTermSet anzsrc-for:
127 schema:name Information and Computing Sciences
128 rdf:type schema:DefinedTerm
129 anzsrc-for:0801 schema:inDefinedTermSet anzsrc-for:
130 schema:name Artificial Intelligence and Image Processing
131 rdf:type schema:DefinedTerm
132 sg:person.012303351315.43 schema:affiliation grid-institutes:grid.12527.33
133 schema:familyName Wang
134 schema:givenName Jianmin
135 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012303351315.43
136 rdf:type schema:Person
137 sg:person.013417115303.81 schema:affiliation grid-institutes:grid.12527.33
138 schema:familyName Long
139 schema:givenName Mingsheng
140 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013417115303.81
141 rdf:type schema:Person
142 sg:person.014716677221.46 schema:affiliation grid-institutes:grid.12527.33
143 schema:familyName Wang
144 schema:givenName Ximei
145 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.014716677221.46
146 rdf:type schema:Person
147 sg:person.07422717707.09 schema:affiliation grid-institutes:grid.12527.33
148 schema:familyName Jin
149 schema:givenName Ying
150 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.07422717707.09
151 rdf:type schema:Person
152 grid-institutes:grid.12527.33 schema:alternateName School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China
153 schema:name School of Software, BNRist, Research Center for Big Data, Tsinghua University, Beijing, China
154 rdf:type schema:Organization
 




Preview window. Press ESC to close (or click here)


...