Deep metric learning for sequential data using approximate information

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

Uittreksel

Learning a distance metric provides solutions to many problems where the data exists in a high dimensional space and hand-crafted distance metrics fail to capture its semantical structure. Methods based on deep neural networks such as Siamese or Triplet networks have been developed for learning such metrics. In this paper we present a metric learning method for sequence data based on a RNN-based triplet network. We posit that this model can be trained efficiently with regards to labels by using Jaccard distance as a proxy distance metric. We empirically demonstrate the performance and efficiency of the approach on three different computer log-line datasets.

TaalEngels
TitelMachine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings
RedacteurenPetra Perner
Plaats van productieCham
UitgeverijSpringer
Pagina's269-282
Aantal pagina's14
ISBN van elektronische versie978-3-319-96136-1
ISBN van geprinte versie978-3-319-96135-4
DOI's
StatusGepubliceerd - 8 jul 2018
Evenement14th International Conference on Machine Learning and Data Mining in Pattern Recognition, MLDM 2018 - New York, Verenigde Staten van Amerika
Duur: 15 jul 201819 jul 2018

Publicatie series

NaamLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10934 LNAI
ISSN van geprinte versie0302-9743
ISSN van elektronische versie1611-3349

Congres

Congres14th International Conference on Machine Learning and Data Mining in Pattern Recognition, MLDM 2018
LandVerenigde Staten van Amerika
StadNew York
Periode15/07/1819/07/18

Vingerafdruk

Distance Metric
Labels
Metric
High-dimensional
Neural Networks
Line
Demonstrate
Learning
Deep neural networks
Model

Trefwoorden

    Citeer dit

    Thaler, S., Menkovski, V., & Petkovic, M. (2018). Deep metric learning for sequential data using approximate information. In P. Perner (editor), Machine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings (blz. 269-282). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10934 LNAI). Cham: Springer. DOI: 10.1007/978-3-319-96136-1_22
    Thaler, Stefan ; Menkovski, Vlado ; Petkovic, Milan. / Deep metric learning for sequential data using approximate information. Machine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings. redacteur / Petra Perner. Cham : Springer, 2018. blz. 269-282 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    @inproceedings{1a812ec0bf7f429f9ccd5650a92bff64,
    title = "Deep metric learning for sequential data using approximate information",
    abstract = "Learning a distance metric provides solutions to many problems where the data exists in a high dimensional space and hand-crafted distance metrics fail to capture its semantical structure. Methods based on deep neural networks such as Siamese or Triplet networks have been developed for learning such metrics. In this paper we present a metric learning method for sequence data based on a RNN-based triplet network. We posit that this model can be trained efficiently with regards to labels by using Jaccard distance as a proxy distance metric. We empirically demonstrate the performance and efficiency of the approach on three different computer log-line datasets.",
    keywords = "Deep learning, Efficient metric learning, Triplet network",
    author = "Stefan Thaler and Vlado Menkovski and Milan Petkovic",
    year = "2018",
    month = "7",
    day = "8",
    doi = "10.1007/978-3-319-96136-1_22",
    language = "English",
    isbn = "978-3-319-96135-4",
    series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
    publisher = "Springer",
    pages = "269--282",
    editor = "Petra Perner",
    booktitle = "Machine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings",
    address = "Germany",

    }

    Thaler, S, Menkovski, V & Petkovic, M 2018, Deep metric learning for sequential data using approximate information. in P Perner (redactie), Machine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10934 LNAI, Springer, Cham, blz. 269-282, New York, Verenigde Staten van Amerika, 15/07/18. DOI: 10.1007/978-3-319-96136-1_22

    Deep metric learning for sequential data using approximate information. / Thaler, Stefan; Menkovski, Vlado; Petkovic, Milan.

    Machine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings. redactie / Petra Perner. Cham : Springer, 2018. blz. 269-282 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10934 LNAI).

    Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

    TY - GEN

    T1 - Deep metric learning for sequential data using approximate information

    AU - Thaler,Stefan

    AU - Menkovski,Vlado

    AU - Petkovic,Milan

    PY - 2018/7/8

    Y1 - 2018/7/8

    N2 - Learning a distance metric provides solutions to many problems where the data exists in a high dimensional space and hand-crafted distance metrics fail to capture its semantical structure. Methods based on deep neural networks such as Siamese or Triplet networks have been developed for learning such metrics. In this paper we present a metric learning method for sequence data based on a RNN-based triplet network. We posit that this model can be trained efficiently with regards to labels by using Jaccard distance as a proxy distance metric. We empirically demonstrate the performance and efficiency of the approach on three different computer log-line datasets.

    AB - Learning a distance metric provides solutions to many problems where the data exists in a high dimensional space and hand-crafted distance metrics fail to capture its semantical structure. Methods based on deep neural networks such as Siamese or Triplet networks have been developed for learning such metrics. In this paper we present a metric learning method for sequence data based on a RNN-based triplet network. We posit that this model can be trained efficiently with regards to labels by using Jaccard distance as a proxy distance metric. We empirically demonstrate the performance and efficiency of the approach on three different computer log-line datasets.

    KW - Deep learning

    KW - Efficient metric learning

    KW - Triplet network

    UR - http://www.scopus.com/inward/record.url?scp=85050538775&partnerID=8YFLogxK

    U2 - 10.1007/978-3-319-96136-1_22

    DO - 10.1007/978-3-319-96136-1_22

    M3 - Conference contribution

    SN - 978-3-319-96135-4

    T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

    SP - 269

    EP - 282

    BT - Machine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings

    PB - Springer

    CY - Cham

    ER -

    Thaler S, Menkovski V, Petkovic M. Deep metric learning for sequential data using approximate information. In Perner P, redacteur, Machine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings. Cham: Springer. 2018. blz. 269-282. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Beschikbaar vanaf, DOI: 10.1007/978-3-319-96136-1_22