EM for mixtures View Full Text


Ontology type: schema:ScholarlyArticle     


Article Info

DATE

2015-07

AUTHORS

Jean-Patrick Baudry, Gilles Celeux

ABSTRACT

Maximum likelihood through the EM algorithm is widely used to estimate the parameters in hidden structure models such as Gaussian mixture models. But the EM algorithm has well-documented drawbacks: its solution could be highly dependent from its initial position and it may fail as a result of degeneracies. We stress the practical dangers of theses limitations and how carefully they should be dealt with. Our main conclusion is that no method enables to address them satisfactory in all situations. But improvements are introduced, first, using a penalized log-likelihood of Gaussian mixture models in a Bayesian regularization perspective and, second, choosing the best among several relevant initialisation strategies. In this perspective, we also propose new recursive initialization strategies which prove helpful. They are compared with standard initialization procedures through numerical experiments and their effects on model selection criteria are analyzed. More... »

PAGES

713-726

References to SciGraph publications

Identifiers

URI

http://scigraph.springernature.com/pub.10.1007/s11222-015-9561-x

DOI

http://dx.doi.org/10.1007/s11222-015-9561-x

DIMENSIONS

https://app.dimensions.ai/details/publication/pub.1031728704


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/2008/0104", 
        "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
        "name": "Statistics", 
        "type": "DefinedTerm"
      }, 
      {
        "id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/01", 
        "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
        "name": "Mathematical Sciences", 
        "type": "DefinedTerm"
      }
    ], 
    "author": [
      {
        "affiliation": {
          "alternateName": "Laboratoire de Statistique Th\u00e9orique et Appliqu\u00e9e", 
          "id": "https://www.grid.ac/institutes/grid.463964.a", 
          "name": [
            "Sorbonne Universit\u00e9s, UPMC Univ Paris 06, FRE 3684, LSTA, 75005, Paris, France"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Baudry", 
        "givenName": "Jean-Patrick", 
        "id": "sg:person.016631012445.98", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.016631012445.98"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "D\u00e9partement de Math\u00e9matiques", 
          "id": "https://www.grid.ac/institutes/grid.463900.8", 
          "name": [
            "INRIA-Saclay-\u00cele-de-France, LMO, Universit\u00e9 Paris-Sud, 91405, Orsay, France"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Celeux", 
        "givenName": "Gilles", 
        "id": "sg:person.01143267124.37", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01143267124.37"
        ], 
        "type": "Person"
      }
    ], 
    "citation": [
      {
        "id": "https://doi.org/10.1016/0167-9473(92)90042-e", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1009547665"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1016/0031-3203(94)00125-6", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1018833578"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1016/j.csda.2014.07.005", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1019320478"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1007/s11222-011-9236-1", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1020568508", 
          "https://doi.org/10.1007/s11222-011-9236-1"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1007/s00440-006-0011-8", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1022832323", 
          "https://doi.org/10.1007/s00440-006-0011-8"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1007/s00440-006-0011-8", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1022832323", 
          "https://doi.org/10.1007/s00440-006-0011-8"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1007/s00440-006-0011-8", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1022832323", 
          "https://doi.org/10.1007/s00440-006-0011-8"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1093/bioinformatics/btu845", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1027313689"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1038/nature09715", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1031240109", 
          "https://doi.org/10.1038/nature09715"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1186/1471-2105-12-449", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1032929685", 
          "https://doi.org/10.1186/1471-2105-12-449"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1186/1471-2105-12-449", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1032929685", 
          "https://doi.org/10.1186/1471-2105-12-449"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1007/bf03372103", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1038743704", 
          "https://doi.org/10.1007/bf03372103"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1007/bf03372103", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1038743704", 
          "https://doi.org/10.1007/bf03372103"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1016/s0167-9473(02)00163-9", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1041132081"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1016/s0167-9473(02)00163-9", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1041132081"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1214/aos/1176344136", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1044872629"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1111/1467-9469.00317", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1045828951"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "sg:pub.10.1007/s00357-007-0004-5", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1053439536", 
          "https://doi.org/10.1007/s00357-007-0004-5"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1080/01621459.1997.10474044", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1058305293"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/34.865189", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1061157115"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1198/016214502760047131", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1064198019"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1198/106186005x59603", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1064199497"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.2307/2532201", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1069977629"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1002/9780470191613", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1098661620"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1002/9780470191613", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1098661620"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://app.dimensions.ai/details/publication/pub.1109491899", 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://app.dimensions.ai/details/publication/pub.1109491899", 
        "type": "CreativeWork"
      }
    ], 
    "datePublished": "2015-07", 
    "datePublishedReg": "2015-07-01", 
    "description": "Maximum likelihood through the EM algorithm is widely used to estimate the parameters in hidden structure models such as Gaussian mixture models. But the EM algorithm has well-documented drawbacks: its solution could be highly dependent from its initial position and it may fail as a result of degeneracies. We stress the practical dangers of theses limitations and how carefully they should be dealt with. Our main conclusion is that no method enables to address them satisfactory in all situations. But improvements are introduced, first, using a penalized log-likelihood of Gaussian mixture models in a Bayesian regularization perspective and, second, choosing the best among several relevant initialisation strategies. In this perspective, we also propose new recursive initialization strategies which prove helpful. They are compared with standard initialization procedures through numerical experiments and their effects on model selection criteria are analyzed.", 
    "genre": "research_article", 
    "id": "sg:pub.10.1007/s11222-015-9561-x", 
    "inLanguage": [
      "en"
    ], 
    "isAccessibleForFree": false, 
    "isPartOf": [
      {
        "id": "sg:journal.1327447", 
        "issn": [
          "0960-3174", 
          "1573-1375"
        ], 
        "name": "Statistics and Computing", 
        "type": "Periodical"
      }, 
      {
        "issueNumber": "4", 
        "type": "PublicationIssue"
      }, 
      {
        "type": "PublicationVolume", 
        "volumeNumber": "25"
      }
    ], 
    "name": "EM for mixtures", 
    "pagination": "713-726", 
    "productId": [
      {
        "name": "readcube_id", 
        "type": "PropertyValue", 
        "value": [
          "9725fc1870ba21683df7be876f9723b9059abbd95598a023e01884b9656bb35e"
        ]
      }, 
      {
        "name": "doi", 
        "type": "PropertyValue", 
        "value": [
          "10.1007/s11222-015-9561-x"
        ]
      }, 
      {
        "name": "dimensions_id", 
        "type": "PropertyValue", 
        "value": [
          "pub.1031728704"
        ]
      }
    ], 
    "sameAs": [
      "https://doi.org/10.1007/s11222-015-9561-x", 
      "https://app.dimensions.ai/details/publication/pub.1031728704"
    ], 
    "sdDataset": "articles", 
    "sdDatePublished": "2019-04-11T02:28", 
    "sdLicense": "https://scigraph.springernature.com/explorer/license/", 
    "sdPublisher": {
      "name": "Springer Nature - SN SciGraph project", 
      "type": "Organization"
    }, 
    "sdSource": "s3://com-uberresearch-data-dimensions-target-20181106-alternative/cleanup/v134/2549eaecd7973599484d7c17b260dba0a4ecb94b/merge/v9/a6c9fde33151104705d4d7ff012ea9563521a3ce/jats-lookup/v90/0000000001_0000000264/records_8700_00000589.jsonl", 
    "type": "ScholarlyArticle", 
    "url": "http://link.springer.com/10.1007%2Fs11222-015-9561-x"
  }
]
 

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/s11222-015-9561-x'

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/s11222-015-9561-x'

Turtle is a human-readable linked data format.

curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/s11222-015-9561-x'

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

curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/s11222-015-9561-x'


 

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

136 TRIPLES      21 PREDICATES      47 URIs      19 LITERALS      7 BLANK NODES

Subject Predicate Object
1 sg:pub.10.1007/s11222-015-9561-x schema:about anzsrc-for:01
2 anzsrc-for:0104
3 schema:author N019eb21c37e649c7bb8668d95978dfb2
4 schema:citation sg:pub.10.1007/bf03372103
5 sg:pub.10.1007/s00357-007-0004-5
6 sg:pub.10.1007/s00440-006-0011-8
7 sg:pub.10.1007/s11222-011-9236-1
8 sg:pub.10.1038/nature09715
9 sg:pub.10.1186/1471-2105-12-449
10 https://app.dimensions.ai/details/publication/pub.1109491899
11 https://doi.org/10.1002/9780470191613
12 https://doi.org/10.1016/0031-3203(94)00125-6
13 https://doi.org/10.1016/0167-9473(92)90042-e
14 https://doi.org/10.1016/j.csda.2014.07.005
15 https://doi.org/10.1016/s0167-9473(02)00163-9
16 https://doi.org/10.1080/01621459.1997.10474044
17 https://doi.org/10.1093/bioinformatics/btu845
18 https://doi.org/10.1109/34.865189
19 https://doi.org/10.1111/1467-9469.00317
20 https://doi.org/10.1198/016214502760047131
21 https://doi.org/10.1198/106186005x59603
22 https://doi.org/10.1214/aos/1176344136
23 https://doi.org/10.2307/2532201
24 schema:datePublished 2015-07
25 schema:datePublishedReg 2015-07-01
26 schema:description Maximum likelihood through the EM algorithm is widely used to estimate the parameters in hidden structure models such as Gaussian mixture models. But the EM algorithm has well-documented drawbacks: its solution could be highly dependent from its initial position and it may fail as a result of degeneracies. We stress the practical dangers of theses limitations and how carefully they should be dealt with. Our main conclusion is that no method enables to address them satisfactory in all situations. But improvements are introduced, first, using a penalized log-likelihood of Gaussian mixture models in a Bayesian regularization perspective and, second, choosing the best among several relevant initialisation strategies. In this perspective, we also propose new recursive initialization strategies which prove helpful. They are compared with standard initialization procedures through numerical experiments and their effects on model selection criteria are analyzed.
27 schema:genre research_article
28 schema:inLanguage en
29 schema:isAccessibleForFree false
30 schema:isPartOf N38658324fb4d4e0b96b9d2d191eec9d7
31 N6a912b0a92ac4b4d8f24d54c351cfbb5
32 sg:journal.1327447
33 schema:name EM for mixtures
34 schema:pagination 713-726
35 schema:productId N818d156c56194241b1185d77f085ad9e
36 Ndeff0b038d104befa03713f615346a0a
37 Nf36cf23f73a449c6b26b3b2627f410a2
38 schema:sameAs https://app.dimensions.ai/details/publication/pub.1031728704
39 https://doi.org/10.1007/s11222-015-9561-x
40 schema:sdDatePublished 2019-04-11T02:28
41 schema:sdLicense https://scigraph.springernature.com/explorer/license/
42 schema:sdPublisher N46b43af496bd49e4bd8feb22dc93d481
43 schema:url http://link.springer.com/10.1007%2Fs11222-015-9561-x
44 sgo:license sg:explorer/license/
45 sgo:sdDataset articles
46 rdf:type schema:ScholarlyArticle
47 N019eb21c37e649c7bb8668d95978dfb2 rdf:first sg:person.016631012445.98
48 rdf:rest N826dcd08dab943a2b01b5de9f831f265
49 N38658324fb4d4e0b96b9d2d191eec9d7 schema:issueNumber 4
50 rdf:type schema:PublicationIssue
51 N46b43af496bd49e4bd8feb22dc93d481 schema:name Springer Nature - SN SciGraph project
52 rdf:type schema:Organization
53 N6a912b0a92ac4b4d8f24d54c351cfbb5 schema:volumeNumber 25
54 rdf:type schema:PublicationVolume
55 N818d156c56194241b1185d77f085ad9e schema:name readcube_id
56 schema:value 9725fc1870ba21683df7be876f9723b9059abbd95598a023e01884b9656bb35e
57 rdf:type schema:PropertyValue
58 N826dcd08dab943a2b01b5de9f831f265 rdf:first sg:person.01143267124.37
59 rdf:rest rdf:nil
60 Ndeff0b038d104befa03713f615346a0a schema:name dimensions_id
61 schema:value pub.1031728704
62 rdf:type schema:PropertyValue
63 Nf36cf23f73a449c6b26b3b2627f410a2 schema:name doi
64 schema:value 10.1007/s11222-015-9561-x
65 rdf:type schema:PropertyValue
66 anzsrc-for:01 schema:inDefinedTermSet anzsrc-for:
67 schema:name Mathematical Sciences
68 rdf:type schema:DefinedTerm
69 anzsrc-for:0104 schema:inDefinedTermSet anzsrc-for:
70 schema:name Statistics
71 rdf:type schema:DefinedTerm
72 sg:journal.1327447 schema:issn 0960-3174
73 1573-1375
74 schema:name Statistics and Computing
75 rdf:type schema:Periodical
76 sg:person.01143267124.37 schema:affiliation https://www.grid.ac/institutes/grid.463900.8
77 schema:familyName Celeux
78 schema:givenName Gilles
79 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01143267124.37
80 rdf:type schema:Person
81 sg:person.016631012445.98 schema:affiliation https://www.grid.ac/institutes/grid.463964.a
82 schema:familyName Baudry
83 schema:givenName Jean-Patrick
84 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.016631012445.98
85 rdf:type schema:Person
86 sg:pub.10.1007/bf03372103 schema:sameAs https://app.dimensions.ai/details/publication/pub.1038743704
87 https://doi.org/10.1007/bf03372103
88 rdf:type schema:CreativeWork
89 sg:pub.10.1007/s00357-007-0004-5 schema:sameAs https://app.dimensions.ai/details/publication/pub.1053439536
90 https://doi.org/10.1007/s00357-007-0004-5
91 rdf:type schema:CreativeWork
92 sg:pub.10.1007/s00440-006-0011-8 schema:sameAs https://app.dimensions.ai/details/publication/pub.1022832323
93 https://doi.org/10.1007/s00440-006-0011-8
94 rdf:type schema:CreativeWork
95 sg:pub.10.1007/s11222-011-9236-1 schema:sameAs https://app.dimensions.ai/details/publication/pub.1020568508
96 https://doi.org/10.1007/s11222-011-9236-1
97 rdf:type schema:CreativeWork
98 sg:pub.10.1038/nature09715 schema:sameAs https://app.dimensions.ai/details/publication/pub.1031240109
99 https://doi.org/10.1038/nature09715
100 rdf:type schema:CreativeWork
101 sg:pub.10.1186/1471-2105-12-449 schema:sameAs https://app.dimensions.ai/details/publication/pub.1032929685
102 https://doi.org/10.1186/1471-2105-12-449
103 rdf:type schema:CreativeWork
104 https://app.dimensions.ai/details/publication/pub.1109491899 schema:CreativeWork
105 https://doi.org/10.1002/9780470191613 schema:sameAs https://app.dimensions.ai/details/publication/pub.1098661620
106 rdf:type schema:CreativeWork
107 https://doi.org/10.1016/0031-3203(94)00125-6 schema:sameAs https://app.dimensions.ai/details/publication/pub.1018833578
108 rdf:type schema:CreativeWork
109 https://doi.org/10.1016/0167-9473(92)90042-e schema:sameAs https://app.dimensions.ai/details/publication/pub.1009547665
110 rdf:type schema:CreativeWork
111 https://doi.org/10.1016/j.csda.2014.07.005 schema:sameAs https://app.dimensions.ai/details/publication/pub.1019320478
112 rdf:type schema:CreativeWork
113 https://doi.org/10.1016/s0167-9473(02)00163-9 schema:sameAs https://app.dimensions.ai/details/publication/pub.1041132081
114 rdf:type schema:CreativeWork
115 https://doi.org/10.1080/01621459.1997.10474044 schema:sameAs https://app.dimensions.ai/details/publication/pub.1058305293
116 rdf:type schema:CreativeWork
117 https://doi.org/10.1093/bioinformatics/btu845 schema:sameAs https://app.dimensions.ai/details/publication/pub.1027313689
118 rdf:type schema:CreativeWork
119 https://doi.org/10.1109/34.865189 schema:sameAs https://app.dimensions.ai/details/publication/pub.1061157115
120 rdf:type schema:CreativeWork
121 https://doi.org/10.1111/1467-9469.00317 schema:sameAs https://app.dimensions.ai/details/publication/pub.1045828951
122 rdf:type schema:CreativeWork
123 https://doi.org/10.1198/016214502760047131 schema:sameAs https://app.dimensions.ai/details/publication/pub.1064198019
124 rdf:type schema:CreativeWork
125 https://doi.org/10.1198/106186005x59603 schema:sameAs https://app.dimensions.ai/details/publication/pub.1064199497
126 rdf:type schema:CreativeWork
127 https://doi.org/10.1214/aos/1176344136 schema:sameAs https://app.dimensions.ai/details/publication/pub.1044872629
128 rdf:type schema:CreativeWork
129 https://doi.org/10.2307/2532201 schema:sameAs https://app.dimensions.ai/details/publication/pub.1069977629
130 rdf:type schema:CreativeWork
131 https://www.grid.ac/institutes/grid.463900.8 schema:alternateName Département de Mathématiques
132 schema:name INRIA-Saclay-Île-de-France, LMO, Université Paris-Sud, 91405, Orsay, France
133 rdf:type schema:Organization
134 https://www.grid.ac/institutes/grid.463964.a schema:alternateName Laboratoire de Statistique Théorique et Appliquée
135 schema:name Sorbonne Universités, UPMC Univ Paris 06, FRE 3684, LSTA, 75005, Paris, France
136 rdf:type schema:Organization
 




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


...