Soft Margins for AdaBoost View Full Text


Ontology type: schema:ScholarlyArticle      Open Access: True


Article Info

DATE

2001-03

AUTHORS

G. Rätsch, T. Onoda, K.-R. Müller

ABSTRACT

Recently ensemble methods like ADABOOST have been applied successfully in many problems, while seemingly defying the problems of overfitting. ADABOOST rarely overfits in the low noise regime, however, we show that it clearly does so for higher noise levels. Central to the understanding of this fact is the margin distribution. ADABOOST can be viewed as a constraint gradient descent in an error function with respect to the margin. We find that ADABOOST asymptotically achieves a hard margin distribution, i.e. the algorithm concentrates its resources on a few hard-to-learn patterns that are interestingly very similar to Support Vectors. A hard margin is clearly a sub-optimal strategy in the noisy case, and regularization, in our case a “mistrust” in the data, must be introduced in the algorithm to alleviate the distortions that single difficult patterns (e.g. outliers) can cause to the margin distribution. We propose several regularization methods and generalizations of the original ADABOOST algorithm to achieve a soft margin. In particular we suggest (1) regularized ADABOOSTREG where the gradient decent is done directly with respect to the soft margin and (2) regularized linear and quadratic programming (LP/QP-) ADABOOST, where the soft margin is attained by introducing slack variables. Extensive simulations demonstrate that the proposed regularized ADABOOST-type algorithms are useful and yield competitive results for noisy data. More... »

PAGES

287-320

References to SciGraph publications

  • 1999. Theoretical Views of Boosting in COMPUTATIONAL LEARNING THEORY
  • 1998. An asymptotic analysis of AdaBoost in the binary classification case in ICANN 98
  • 1997. AdaBoosting neural networks: Application to on-line character recognition in ARTIFICIAL NEURAL NETWORKS — ICANN'97
  • 1996. Boosting first-order learning in ALGORITHMIC LEARNING THEORY
  • 1995-09. Support-vector networks in MACHINE LEARNING
  • 1997. A boosting algorithm for regression in ARTIFICIAL NEURAL NETWORKS — ICANN'97
  • 1984-03. Optimization by simulated annealing: Quantitative studies in JOURNAL OF STATISTICAL PHYSICS
  • 1996-08. Bagging predictors in MACHINE LEARNING
  • 2000-03. Improved Generalization Through Explicit Optimization of Margins in MACHINE LEARNING
  • Identifiers

    URI

    http://scigraph.springernature.com/pub.10.1023/a:1007618119488

    DOI

    http://dx.doi.org/10.1023/a:1007618119488

    DIMENSIONS

    https://app.dimensions.ai/details/publication/pub.1003090683


    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/2008/0801", 
            "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
            "name": "Artificial Intelligence and Image Processing", 
            "type": "DefinedTerm"
          }, 
          {
            "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"
          }
        ], 
        "author": [
          {
            "affiliation": {
              "name": [
                "GMD FIRST, Kekul\u00e9str. 7, 12489, Berlin, Germany"
              ], 
              "type": "Organization"
            }, 
            "familyName": "R\u00e4tsch", 
            "givenName": "G.", 
            "id": "sg:person.01005551566.12", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01005551566.12"
            ], 
            "type": "Person"
          }, 
          {
            "affiliation": {
              "alternateName": "Central Research Institute of Electric Power Industry", 
              "id": "https://www.grid.ac/institutes/grid.417751.1", 
              "name": [
                "CRIEPI, Komae-shi, 2-11-1, Iwado Kita, Tokyo, Japan"
              ], 
              "type": "Organization"
            }, 
            "familyName": "Onoda", 
            "givenName": "T.", 
            "id": "sg:person.013134307141.22", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013134307141.22"
            ], 
            "type": "Person"
          }, 
          {
            "affiliation": {
              "alternateName": "University of Potsdam", 
              "id": "https://www.grid.ac/institutes/grid.11348.3f", 
              "name": [
                "GMD FIRST, Kekul\u00e9str. 7, 12489, Berlin, Germany", 
                "University of Potsdam, Neues Palais 10, 14469, Potsdam, Germany"
              ], 
              "type": "Organization"
            }, 
            "familyName": "M\u00fcller", 
            "givenName": "K.-R.", 
            "id": "sg:person.01066211757.29", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01066211757.29"
            ], 
            "type": "Person"
          }
        ], 
        "citation": [
          {
            "id": "sg:pub.10.1007/978-1-4471-1599-1_26", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1000123091", 
              "https://doi.org/10.1007/978-1-4471-1599-1_26"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/bf00058655", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1002929950", 
              "https://doi.org/10.1007/bf00058655"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1080/10556789208805504", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1006237986"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1162/neco.1989.1.2.281", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1018769097"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/bf01009452", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1019054075", 
              "https://doi.org/10.1007/bf01009452"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/3-540-49097-3_1", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1019257929", 
              "https://doi.org/10.1007/3-540-49097-3_1"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1016/s0893-6080(98)00032-x", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1021275024"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/bf00994018", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1025150743", 
              "https://doi.org/10.1007/bf00994018"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1145/238061.238163", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1027584775"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/3-540-61863-5_42", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1029474067", 
              "https://doi.org/10.1007/3-540-61863-5_42"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1162/089976600300015565", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1031092269"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1023/a:1007697429651", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1033223233", 
              "https://doi.org/10.1023/a:1007697429651"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1214/aos/1024691352", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1035391848"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1145/130385.130401", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1036379424"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/bfb0020278", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1040719502", 
              "https://doi.org/10.1007/bfb0020278"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/bfb0020178", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1045162027", 
              "https://doi.org/10.1007/bfb0020178"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1287/opre.13.3.444", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1064726865"
            ], 
            "type": "CreativeWork"
          }
        ], 
        "datePublished": "2001-03", 
        "datePublishedReg": "2001-03-01", 
        "description": "Recently ensemble methods like ADABOOST have been applied successfully in many problems, while seemingly defying the problems of overfitting. ADABOOST rarely overfits in the low noise regime, however, we show that it clearly does so for higher noise levels. Central to the understanding of this fact is the margin distribution. ADABOOST can be viewed as a constraint gradient descent in an error function with respect to the margin. We find that ADABOOST asymptotically achieves a hard margin distribution, i.e. the algorithm concentrates its resources on a few hard-to-learn patterns that are interestingly very similar to Support Vectors. A hard margin is clearly a sub-optimal strategy in the noisy case, and regularization, in our case a \u201cmistrust\u201d in the data, must be introduced in the algorithm to alleviate the distortions that single difficult patterns (e.g. outliers) can cause to the margin distribution. We propose several regularization methods and generalizations of the original ADABOOST algorithm to achieve a soft margin. In particular we suggest (1) regularized ADABOOSTREG where the gradient decent is done directly with respect to the soft margin and (2) regularized linear and quadratic programming (LP/QP-) ADABOOST, where the soft margin is attained by introducing slack variables. Extensive simulations demonstrate that the proposed regularized ADABOOST-type algorithms are useful and yield competitive results for noisy data.", 
        "genre": "research_article", 
        "id": "sg:pub.10.1023/a:1007618119488", 
        "inLanguage": [
          "en"
        ], 
        "isAccessibleForFree": true, 
        "isPartOf": [
          {
            "id": "sg:journal.1125588", 
            "issn": [
              "0885-6125", 
              "1573-0565"
            ], 
            "name": "Machine Learning", 
            "type": "Periodical"
          }, 
          {
            "issueNumber": "3", 
            "type": "PublicationIssue"
          }, 
          {
            "type": "PublicationVolume", 
            "volumeNumber": "42"
          }
        ], 
        "name": "Soft Margins for AdaBoost", 
        "pagination": "287-320", 
        "productId": [
          {
            "name": "readcube_id", 
            "type": "PropertyValue", 
            "value": [
              "0eb912ebbb58f9aa19d82e6223f5bfb280aafece306bf56fbd6676185e8741b1"
            ]
          }, 
          {
            "name": "doi", 
            "type": "PropertyValue", 
            "value": [
              "10.1023/a:1007618119488"
            ]
          }, 
          {
            "name": "dimensions_id", 
            "type": "PropertyValue", 
            "value": [
              "pub.1003090683"
            ]
          }
        ], 
        "sameAs": [
          "https://doi.org/10.1023/a:1007618119488", 
          "https://app.dimensions.ai/details/publication/pub.1003090683"
        ], 
        "sdDataset": "articles", 
        "sdDatePublished": "2019-04-10T19:54", 
        "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_8681_00000498.jsonl", 
        "type": "ScholarlyArticle", 
        "url": "http://link.springer.com/10.1023/A:1007618119488"
      }
    ]
     

    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.10.1023/a:1007618119488'

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

    curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1023/a:1007618119488'

    Turtle is a human-readable linked data format.

    curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1023/a:1007618119488'

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

    curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1023/a:1007618119488'


     

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

    141 TRIPLES      21 PREDICATES      44 URIs      19 LITERALS      7 BLANK NODES

    Subject Predicate Object
    1 sg:pub.10.1023/a:1007618119488 schema:about anzsrc-for:08
    2 anzsrc-for:0801
    3 schema:author N153b7ea5bcee48d4a440a3edb4aeb77e
    4 schema:citation sg:pub.10.1007/3-540-49097-3_1
    5 sg:pub.10.1007/3-540-61863-5_42
    6 sg:pub.10.1007/978-1-4471-1599-1_26
    7 sg:pub.10.1007/bf00058655
    8 sg:pub.10.1007/bf00994018
    9 sg:pub.10.1007/bf01009452
    10 sg:pub.10.1007/bfb0020178
    11 sg:pub.10.1007/bfb0020278
    12 sg:pub.10.1023/a:1007697429651
    13 https://doi.org/10.1016/s0893-6080(98)00032-x
    14 https://doi.org/10.1080/10556789208805504
    15 https://doi.org/10.1145/130385.130401
    16 https://doi.org/10.1145/238061.238163
    17 https://doi.org/10.1162/089976600300015565
    18 https://doi.org/10.1162/neco.1989.1.2.281
    19 https://doi.org/10.1214/aos/1024691352
    20 https://doi.org/10.1287/opre.13.3.444
    21 schema:datePublished 2001-03
    22 schema:datePublishedReg 2001-03-01
    23 schema:description Recently ensemble methods like ADABOOST have been applied successfully in many problems, while seemingly defying the problems of overfitting. ADABOOST rarely overfits in the low noise regime, however, we show that it clearly does so for higher noise levels. Central to the understanding of this fact is the margin distribution. ADABOOST can be viewed as a constraint gradient descent in an error function with respect to the margin. We find that ADABOOST asymptotically achieves a hard margin distribution, i.e. the algorithm concentrates its resources on a few hard-to-learn patterns that are interestingly very similar to Support Vectors. A hard margin is clearly a sub-optimal strategy in the noisy case, and regularization, in our case a “mistrust” in the data, must be introduced in the algorithm to alleviate the distortions that single difficult patterns (e.g. outliers) can cause to the margin distribution. We propose several regularization methods and generalizations of the original ADABOOST algorithm to achieve a soft margin. In particular we suggest (1) regularized ADABOOSTREG where the gradient decent is done directly with respect to the soft margin and (2) regularized linear and quadratic programming (LP/QP-) ADABOOST, where the soft margin is attained by introducing slack variables. Extensive simulations demonstrate that the proposed regularized ADABOOST-type algorithms are useful and yield competitive results for noisy data.
    24 schema:genre research_article
    25 schema:inLanguage en
    26 schema:isAccessibleForFree true
    27 schema:isPartOf N219b386c758c4f299189813326d01f67
    28 N78d89498ffeb4807b0372a9c8f4439c1
    29 sg:journal.1125588
    30 schema:name Soft Margins for AdaBoost
    31 schema:pagination 287-320
    32 schema:productId N346cc8cec17d40eea4ef18ce9b6c0c02
    33 N4c495bb970b84a8c95540f836a9a16b7
    34 Nd50a2f31971b44329ef09f3bf875232e
    35 schema:sameAs https://app.dimensions.ai/details/publication/pub.1003090683
    36 https://doi.org/10.1023/a:1007618119488
    37 schema:sdDatePublished 2019-04-10T19:54
    38 schema:sdLicense https://scigraph.springernature.com/explorer/license/
    39 schema:sdPublisher N289ed9d004f74db18b5155e77a2704c0
    40 schema:url http://link.springer.com/10.1023/A:1007618119488
    41 sgo:license sg:explorer/license/
    42 sgo:sdDataset articles
    43 rdf:type schema:ScholarlyArticle
    44 N153b7ea5bcee48d4a440a3edb4aeb77e rdf:first sg:person.01005551566.12
    45 rdf:rest N50d7b4816d004f01a0a6a22fee85db29
    46 N219b386c758c4f299189813326d01f67 schema:issueNumber 3
    47 rdf:type schema:PublicationIssue
    48 N24d79b42d37a4ad0a5ad54972e0925ee schema:name GMD FIRST, Kekuléstr. 7, 12489, Berlin, Germany
    49 rdf:type schema:Organization
    50 N289ed9d004f74db18b5155e77a2704c0 schema:name Springer Nature - SN SciGraph project
    51 rdf:type schema:Organization
    52 N346cc8cec17d40eea4ef18ce9b6c0c02 schema:name doi
    53 schema:value 10.1023/a:1007618119488
    54 rdf:type schema:PropertyValue
    55 N4c495bb970b84a8c95540f836a9a16b7 schema:name readcube_id
    56 schema:value 0eb912ebbb58f9aa19d82e6223f5bfb280aafece306bf56fbd6676185e8741b1
    57 rdf:type schema:PropertyValue
    58 N50d7b4816d004f01a0a6a22fee85db29 rdf:first sg:person.013134307141.22
    59 rdf:rest Na2fead81c5004c6eac1a4f77fd0f40f4
    60 N78d89498ffeb4807b0372a9c8f4439c1 schema:volumeNumber 42
    61 rdf:type schema:PublicationVolume
    62 Na2fead81c5004c6eac1a4f77fd0f40f4 rdf:first sg:person.01066211757.29
    63 rdf:rest rdf:nil
    64 Nd50a2f31971b44329ef09f3bf875232e schema:name dimensions_id
    65 schema:value pub.1003090683
    66 rdf:type schema:PropertyValue
    67 anzsrc-for:08 schema:inDefinedTermSet anzsrc-for:
    68 schema:name Information and Computing Sciences
    69 rdf:type schema:DefinedTerm
    70 anzsrc-for:0801 schema:inDefinedTermSet anzsrc-for:
    71 schema:name Artificial Intelligence and Image Processing
    72 rdf:type schema:DefinedTerm
    73 sg:journal.1125588 schema:issn 0885-6125
    74 1573-0565
    75 schema:name Machine Learning
    76 rdf:type schema:Periodical
    77 sg:person.01005551566.12 schema:affiliation N24d79b42d37a4ad0a5ad54972e0925ee
    78 schema:familyName Rätsch
    79 schema:givenName G.
    80 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01005551566.12
    81 rdf:type schema:Person
    82 sg:person.01066211757.29 schema:affiliation https://www.grid.ac/institutes/grid.11348.3f
    83 schema:familyName Müller
    84 schema:givenName K.-R.
    85 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01066211757.29
    86 rdf:type schema:Person
    87 sg:person.013134307141.22 schema:affiliation https://www.grid.ac/institutes/grid.417751.1
    88 schema:familyName Onoda
    89 schema:givenName T.
    90 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013134307141.22
    91 rdf:type schema:Person
    92 sg:pub.10.1007/3-540-49097-3_1 schema:sameAs https://app.dimensions.ai/details/publication/pub.1019257929
    93 https://doi.org/10.1007/3-540-49097-3_1
    94 rdf:type schema:CreativeWork
    95 sg:pub.10.1007/3-540-61863-5_42 schema:sameAs https://app.dimensions.ai/details/publication/pub.1029474067
    96 https://doi.org/10.1007/3-540-61863-5_42
    97 rdf:type schema:CreativeWork
    98 sg:pub.10.1007/978-1-4471-1599-1_26 schema:sameAs https://app.dimensions.ai/details/publication/pub.1000123091
    99 https://doi.org/10.1007/978-1-4471-1599-1_26
    100 rdf:type schema:CreativeWork
    101 sg:pub.10.1007/bf00058655 schema:sameAs https://app.dimensions.ai/details/publication/pub.1002929950
    102 https://doi.org/10.1007/bf00058655
    103 rdf:type schema:CreativeWork
    104 sg:pub.10.1007/bf00994018 schema:sameAs https://app.dimensions.ai/details/publication/pub.1025150743
    105 https://doi.org/10.1007/bf00994018
    106 rdf:type schema:CreativeWork
    107 sg:pub.10.1007/bf01009452 schema:sameAs https://app.dimensions.ai/details/publication/pub.1019054075
    108 https://doi.org/10.1007/bf01009452
    109 rdf:type schema:CreativeWork
    110 sg:pub.10.1007/bfb0020178 schema:sameAs https://app.dimensions.ai/details/publication/pub.1045162027
    111 https://doi.org/10.1007/bfb0020178
    112 rdf:type schema:CreativeWork
    113 sg:pub.10.1007/bfb0020278 schema:sameAs https://app.dimensions.ai/details/publication/pub.1040719502
    114 https://doi.org/10.1007/bfb0020278
    115 rdf:type schema:CreativeWork
    116 sg:pub.10.1023/a:1007697429651 schema:sameAs https://app.dimensions.ai/details/publication/pub.1033223233
    117 https://doi.org/10.1023/a:1007697429651
    118 rdf:type schema:CreativeWork
    119 https://doi.org/10.1016/s0893-6080(98)00032-x schema:sameAs https://app.dimensions.ai/details/publication/pub.1021275024
    120 rdf:type schema:CreativeWork
    121 https://doi.org/10.1080/10556789208805504 schema:sameAs https://app.dimensions.ai/details/publication/pub.1006237986
    122 rdf:type schema:CreativeWork
    123 https://doi.org/10.1145/130385.130401 schema:sameAs https://app.dimensions.ai/details/publication/pub.1036379424
    124 rdf:type schema:CreativeWork
    125 https://doi.org/10.1145/238061.238163 schema:sameAs https://app.dimensions.ai/details/publication/pub.1027584775
    126 rdf:type schema:CreativeWork
    127 https://doi.org/10.1162/089976600300015565 schema:sameAs https://app.dimensions.ai/details/publication/pub.1031092269
    128 rdf:type schema:CreativeWork
    129 https://doi.org/10.1162/neco.1989.1.2.281 schema:sameAs https://app.dimensions.ai/details/publication/pub.1018769097
    130 rdf:type schema:CreativeWork
    131 https://doi.org/10.1214/aos/1024691352 schema:sameAs https://app.dimensions.ai/details/publication/pub.1035391848
    132 rdf:type schema:CreativeWork
    133 https://doi.org/10.1287/opre.13.3.444 schema:sameAs https://app.dimensions.ai/details/publication/pub.1064726865
    134 rdf:type schema:CreativeWork
    135 https://www.grid.ac/institutes/grid.11348.3f schema:alternateName University of Potsdam
    136 schema:name GMD FIRST, Kekuléstr. 7, 12489, Berlin, Germany
    137 University of Potsdam, Neues Palais 10, 14469, Potsdam, Germany
    138 rdf:type schema:Organization
    139 https://www.grid.ac/institutes/grid.417751.1 schema:alternateName Central Research Institute of Electric Power Industry
    140 schema:name CRIEPI, Komae-shi, 2-11-1, Iwado Kita, Tokyo, Japan
    141 rdf:type schema:Organization
     




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


    ...