Fighting Class Imbalance with Contrastive Learning View Full Text


Ontology type: schema:Chapter     


Chapter Info

DATE

2021-09-21

AUTHORS

Yassine Marrakchi , Osama Makansi , Thomas Brox

ABSTRACT

Medical image datasets are hard to collect, expensive to label, and often highly imbalanced. The last issue is underestimated, as typical average metrics hardly reveal that the often very important minority classes have a very low accuracy. In this paper, we address this problem by a feature embedding that balances the classes using contrastive learning as an alternative to the common cross-entropy loss. The approach is largely orthogonal to existing sampling methods and can be easily combined with those. We show on the challenging ISIC2018 and APTOS2019 datasets that the approach improves especially the accuracy of minority classes without negatively affecting the majority ones. More... »

PAGES

466-476

Identifiers

URI

http://scigraph.springernature.com/pub.10.1007/978-3-030-87199-4_44

DOI

http://dx.doi.org/10.1007/978-3-030-87199-4_44

DIMENSIONS

https://app.dimensions.ai/details/publication/pub.1141326834


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": "CIBSS \u2013 Centre for Integrative Biological Signalling Studies, University of Freiburg, Freiburg im Breisgau, Germany", 
          "id": "http://www.grid.ac/institutes/grid.5963.9", 
          "name": [
            "Department of Computer Science, University of Freiburg, Freiburg im Breisgau, Germany", 
            "CIBSS \u2013 Centre for Integrative Biological Signalling Studies, University of Freiburg, Freiburg im Breisgau, Germany"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Marrakchi", 
        "givenName": "Yassine", 
        "id": "sg:person.014511262624.07", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.014511262624.07"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "Department of Computer Science, University of Freiburg, Freiburg im Breisgau, Germany", 
          "id": "http://www.grid.ac/institutes/grid.5963.9", 
          "name": [
            "Department of Computer Science, University of Freiburg, Freiburg im Breisgau, Germany"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Makansi", 
        "givenName": "Osama", 
        "id": "sg:person.013221150447.86", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013221150447.86"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "CIBSS \u2013 Centre for Integrative Biological Signalling Studies, University of Freiburg, Freiburg im Breisgau, Germany", 
          "id": "http://www.grid.ac/institutes/grid.5963.9", 
          "name": [
            "Department of Computer Science, University of Freiburg, Freiburg im Breisgau, Germany", 
            "CIBSS \u2013 Centre for Integrative Biological Signalling Studies, University of Freiburg, Freiburg im Breisgau, Germany"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Brox", 
        "givenName": "Thomas", 
        "id": "sg:person.012443225372.65", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012443225372.65"
        ], 
        "type": "Person"
      }
    ], 
    "datePublished": "2021-09-21", 
    "datePublishedReg": "2021-09-21", 
    "description": "Medical image datasets are hard to collect, expensive to label, and often highly imbalanced. The last issue is underestimated, as typical average metrics hardly reveal that the often very important minority classes have a very low accuracy. In this paper, we address this problem by a feature embedding that balances the classes using contrastive learning as an alternative to the common cross-entropy loss. The approach is largely orthogonal to existing sampling methods and can be easily combined with those. We show on the challenging ISIC2018 and APTOS2019 datasets that the approach improves especially the accuracy of minority classes without negatively affecting the majority ones.", 
    "editor": [
      {
        "familyName": "de Bruijne", 
        "givenName": "Marleen", 
        "type": "Person"
      }, 
      {
        "familyName": "Cattin", 
        "givenName": "Philippe C.", 
        "type": "Person"
      }, 
      {
        "familyName": "Cotin", 
        "givenName": "St\u00e9phane", 
        "type": "Person"
      }, 
      {
        "familyName": "Padoy", 
        "givenName": "Nicolas", 
        "type": "Person"
      }, 
      {
        "familyName": "Speidel", 
        "givenName": "Stefanie", 
        "type": "Person"
      }, 
      {
        "familyName": "Zheng", 
        "givenName": "Yefeng", 
        "type": "Person"
      }, 
      {
        "familyName": "Essert", 
        "givenName": "Caroline", 
        "type": "Person"
      }
    ], 
    "genre": "chapter", 
    "id": "sg:pub.10.1007/978-3-030-87199-4_44", 
    "isAccessibleForFree": false, 
    "isPartOf": {
      "isbn": [
        "978-3-030-87198-7", 
        "978-3-030-87199-4"
      ], 
      "name": "Medical Image Computing and Computer Assisted Intervention \u2013 MICCAI 2021", 
      "type": "Book"
    }, 
    "keywords": [
      "low accuracy", 
      "accuracy", 
      "image datasets", 
      "approach", 
      "important minority class", 
      "method", 
      "medical image datasets", 
      "ISIC2018", 
      "alternative", 
      "problem", 
      "majority one", 
      "loss", 
      "one", 
      "last issue", 
      "issues", 
      "cross-entropy loss", 
      "dataset", 
      "metrics", 
      "class", 
      "sampling method", 
      "average metrics", 
      "embedding", 
      "imbalance", 
      "feature embedding", 
      "class imbalance", 
      "contrastive learning", 
      "minority class", 
      "learning", 
      "paper"
    ], 
    "name": "Fighting Class Imbalance with Contrastive Learning", 
    "pagination": "466-476", 
    "productId": [
      {
        "name": "dimensions_id", 
        "type": "PropertyValue", 
        "value": [
          "pub.1141326834"
        ]
      }, 
      {
        "name": "doi", 
        "type": "PropertyValue", 
        "value": [
          "10.1007/978-3-030-87199-4_44"
        ]
      }
    ], 
    "publisher": {
      "name": "Springer Nature", 
      "type": "Organisation"
    }, 
    "sameAs": [
      "https://doi.org/10.1007/978-3-030-87199-4_44", 
      "https://app.dimensions.ai/details/publication/pub.1141326834"
    ], 
    "sdDataset": "chapters", 
    "sdDatePublished": "2022-10-01T06:56", 
    "sdLicense": "https://scigraph.springernature.com/explorer/license/", 
    "sdPublisher": {
      "name": "Springer Nature - SN SciGraph project", 
      "type": "Organization"
    }, 
    "sdSource": "s3://com-springernature-scigraph/baseset/20221001/entities/gbq_results/chapter/chapter_32.jsonl", 
    "type": "Chapter", 
    "url": "https://doi.org/10.1007/978-3-030-87199-4_44"
  }
]
 

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.10.1007/978-3-030-87199-4_44'

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

curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-87199-4_44'

Turtle is a human-readable linked data format.

curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-87199-4_44'

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

curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-87199-4_44'


 

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

134 TRIPLES      22 PREDICATES      53 URIs      46 LITERALS      7 BLANK NODES

Subject Predicate Object
1 sg:pub.10.1007/978-3-030-87199-4_44 schema:about anzsrc-for:08
2 anzsrc-for:0801
3 schema:author Nd5982ab08f3d4c7bbcbd02a54a2ac5d4
4 schema:datePublished 2021-09-21
5 schema:datePublishedReg 2021-09-21
6 schema:description Medical image datasets are hard to collect, expensive to label, and often highly imbalanced. The last issue is underestimated, as typical average metrics hardly reveal that the often very important minority classes have a very low accuracy. In this paper, we address this problem by a feature embedding that balances the classes using contrastive learning as an alternative to the common cross-entropy loss. The approach is largely orthogonal to existing sampling methods and can be easily combined with those. We show on the challenging ISIC2018 and APTOS2019 datasets that the approach improves especially the accuracy of minority classes without negatively affecting the majority ones.
7 schema:editor N26042a8b5e444e22b258aaddab7a035d
8 schema:genre chapter
9 schema:isAccessibleForFree false
10 schema:isPartOf N7c14c15725a14f6bb474dbc873162f0d
11 schema:keywords ISIC2018
12 accuracy
13 alternative
14 approach
15 average metrics
16 class
17 class imbalance
18 contrastive learning
19 cross-entropy loss
20 dataset
21 embedding
22 feature embedding
23 image datasets
24 imbalance
25 important minority class
26 issues
27 last issue
28 learning
29 loss
30 low accuracy
31 majority one
32 medical image datasets
33 method
34 metrics
35 minority class
36 one
37 paper
38 problem
39 sampling method
40 schema:name Fighting Class Imbalance with Contrastive Learning
41 schema:pagination 466-476
42 schema:productId Na3209acd757a45e7a1e551473a37b145
43 Nccf315a6301c4c93af19cab46843e6f2
44 schema:publisher Ne3746e5493664b6899fcbd70db2a734a
45 schema:sameAs https://app.dimensions.ai/details/publication/pub.1141326834
46 https://doi.org/10.1007/978-3-030-87199-4_44
47 schema:sdDatePublished 2022-10-01T06:56
48 schema:sdLicense https://scigraph.springernature.com/explorer/license/
49 schema:sdPublisher N1a00d46b882149e6bd8ff583f7167b5d
50 schema:url https://doi.org/10.1007/978-3-030-87199-4_44
51 sgo:license sg:explorer/license/
52 sgo:sdDataset chapters
53 rdf:type schema:Chapter
54 N0261efb9016d4767821b1be95179882e schema:familyName Padoy
55 schema:givenName Nicolas
56 rdf:type schema:Person
57 N07e5d4bff6d04b4b807c2adcb26f9cba schema:familyName Cotin
58 schema:givenName Stéphane
59 rdf:type schema:Person
60 N1a00d46b882149e6bd8ff583f7167b5d schema:name Springer Nature - SN SciGraph project
61 rdf:type schema:Organization
62 N26042a8b5e444e22b258aaddab7a035d rdf:first Nd393468bad294004971580da9cac46a6
63 rdf:rest N747825ea58594d11b90ba475318f8666
64 N2d9710dbfdf8494e83444f24843f70eb rdf:first N07e5d4bff6d04b4b807c2adcb26f9cba
65 rdf:rest N4ca0330e761043fa9ee60ad69c4aaa87
66 N2f22432651964ac7afd8f2b4a51cd09c rdf:first Nde7658d363c24de9ba0e6e37652e32f9
67 rdf:rest rdf:nil
68 N30a4e5ae79f84131a887fae06b8107e7 rdf:first Ne358d7328c68411eaafd87462f75aa3e
69 rdf:rest N6d3ce19790f8430a96456027ee66105c
70 N3eac89ff084a480bb6d36b71dc830e31 schema:familyName Zheng
71 schema:givenName Yefeng
72 rdf:type schema:Person
73 N4ca0330e761043fa9ee60ad69c4aaa87 rdf:first N0261efb9016d4767821b1be95179882e
74 rdf:rest N30a4e5ae79f84131a887fae06b8107e7
75 N6338e872128f462e885245e98189b863 rdf:first sg:person.012443225372.65
76 rdf:rest rdf:nil
77 N695d7317b9f64b4693da51ff5e97962c schema:familyName Cattin
78 schema:givenName Philippe C.
79 rdf:type schema:Person
80 N6d3ce19790f8430a96456027ee66105c rdf:first N3eac89ff084a480bb6d36b71dc830e31
81 rdf:rest N2f22432651964ac7afd8f2b4a51cd09c
82 N747825ea58594d11b90ba475318f8666 rdf:first N695d7317b9f64b4693da51ff5e97962c
83 rdf:rest N2d9710dbfdf8494e83444f24843f70eb
84 N7c14c15725a14f6bb474dbc873162f0d schema:isbn 978-3-030-87198-7
85 978-3-030-87199-4
86 schema:name Medical Image Computing and Computer Assisted Intervention – MICCAI 2021
87 rdf:type schema:Book
88 N9d82a618c13c4ea6b65ce21649dd886e rdf:first sg:person.013221150447.86
89 rdf:rest N6338e872128f462e885245e98189b863
90 Na3209acd757a45e7a1e551473a37b145 schema:name doi
91 schema:value 10.1007/978-3-030-87199-4_44
92 rdf:type schema:PropertyValue
93 Nccf315a6301c4c93af19cab46843e6f2 schema:name dimensions_id
94 schema:value pub.1141326834
95 rdf:type schema:PropertyValue
96 Nd393468bad294004971580da9cac46a6 schema:familyName de Bruijne
97 schema:givenName Marleen
98 rdf:type schema:Person
99 Nd5982ab08f3d4c7bbcbd02a54a2ac5d4 rdf:first sg:person.014511262624.07
100 rdf:rest N9d82a618c13c4ea6b65ce21649dd886e
101 Nde7658d363c24de9ba0e6e37652e32f9 schema:familyName Essert
102 schema:givenName Caroline
103 rdf:type schema:Person
104 Ne358d7328c68411eaafd87462f75aa3e schema:familyName Speidel
105 schema:givenName Stefanie
106 rdf:type schema:Person
107 Ne3746e5493664b6899fcbd70db2a734a schema:name Springer Nature
108 rdf:type schema:Organisation
109 anzsrc-for:08 schema:inDefinedTermSet anzsrc-for:
110 schema:name Information and Computing Sciences
111 rdf:type schema:DefinedTerm
112 anzsrc-for:0801 schema:inDefinedTermSet anzsrc-for:
113 schema:name Artificial Intelligence and Image Processing
114 rdf:type schema:DefinedTerm
115 sg:person.012443225372.65 schema:affiliation grid-institutes:grid.5963.9
116 schema:familyName Brox
117 schema:givenName Thomas
118 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012443225372.65
119 rdf:type schema:Person
120 sg:person.013221150447.86 schema:affiliation grid-institutes:grid.5963.9
121 schema:familyName Makansi
122 schema:givenName Osama
123 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013221150447.86
124 rdf:type schema:Person
125 sg:person.014511262624.07 schema:affiliation grid-institutes:grid.5963.9
126 schema:familyName Marrakchi
127 schema:givenName Yassine
128 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.014511262624.07
129 rdf:type schema:Person
130 grid-institutes:grid.5963.9 schema:alternateName CIBSS – Centre for Integrative Biological Signalling Studies, University of Freiburg, Freiburg im Breisgau, Germany
131 Department of Computer Science, University of Freiburg, Freiburg im Breisgau, Germany
132 schema:name CIBSS – Centre for Integrative Biological Signalling Studies, University of Freiburg, Freiburg im Breisgau, Germany
133 Department of Computer Science, University of Freiburg, Freiburg im Breisgau, Germany
134 rdf:type schema:Organization
 




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


...