Conference paper Open Access

INPUT COMPLEXITY AND OUT-OF-DISTRIBUTION DETECTION WITH LIKELIHOOD-BASED GENERATIVE MODELS

Serra, Joan; Álvarez, David; Gómez, Vicenç; Slizovskaia , Olga; F. Núñez, José; Luque, Jordi


MARC21 XML Export

<?xml version='1.0' encoding='UTF-8'?>
<record xmlns="http://www.loc.gov/MARC21/slim">
  <leader>00000nam##2200000uu#4500</leader>
  <controlfield tag="005">20210825134824.0</controlfield>
  <datafield tag="500" ind1=" " ind2=" ">
    <subfield code="a">Accepted for ICLR2020</subfield>
  </datafield>
  <controlfield tag="001">5253740</controlfield>
  <datafield tag="711" ind1=" " ind2=" ">
    <subfield code="d">May 3rd through the 7th, September, 2021</subfield>
    <subfield code="g">ICLR 2020</subfield>
    <subfield code="a">Ninth International Conference on Learning Representation</subfield>
    <subfield code="c">Addiss Abbeba</subfield>
    <subfield code="n">https://iclr.cc/virtual_2020/poster_SyxIWpVYvr.html</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">Universitat Politècnica de Catalunya,  Barcelona, Spain</subfield>
    <subfield code="a">Álvarez, David</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">Universitat Pompeu Fabra, Barcelona, Spain</subfield>
    <subfield code="a">Gómez, Vicenç</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">Universitat Pompeu Fabra, Barcelona, Spain</subfield>
    <subfield code="a">Slizovskaia , Olga</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">Universitat Pompeu Fabra, Barcelona, Spain</subfield>
    <subfield code="a">F. Núñez, José</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">Telefónica I+D, Research, Barcelona, Spain</subfield>
    <subfield code="a">Luque, Jordi</subfield>
  </datafield>
  <datafield tag="856" ind1="4" ind2=" ">
    <subfield code="s">1941308</subfield>
    <subfield code="z">md5:fa030c9fd1c34044adfdcce3c9cbcdb8</subfield>
    <subfield code="u">https://zenodo.org/record/5253740/files/input_complexity_and_out_of_di.pdf</subfield>
  </datafield>
  <datafield tag="542" ind1=" " ind2=" ">
    <subfield code="l">open</subfield>
  </datafield>
  <datafield tag="856" ind1="4" ind2=" ">
    <subfield code="y">Conference website</subfield>
    <subfield code="u">https://iclr.cc</subfield>
  </datafield>
  <datafield tag="260" ind1=" " ind2=" ">
    <subfield code="c">2020-04-30</subfield>
  </datafield>
  <datafield tag="909" ind1="C" ind2="O">
    <subfield code="p">openaire</subfield>
    <subfield code="p">user-accordion</subfield>
    <subfield code="o">oai:zenodo.org:5253740</subfield>
  </datafield>
  <datafield tag="100" ind1=" " ind2=" ">
    <subfield code="u">Dolby Laboratories,  Barcelona, Spain</subfield>
    <subfield code="a">Serra, Joan</subfield>
  </datafield>
  <datafield tag="245" ind1=" " ind2=" ">
    <subfield code="a">INPUT COMPLEXITY AND OUT-OF-DISTRIBUTION DETECTION WITH LIKELIHOOD-BASED GENERATIVE MODELS</subfield>
  </datafield>
  <datafield tag="980" ind1=" " ind2=" ">
    <subfield code="a">user-accordion</subfield>
  </datafield>
  <datafield tag="536" ind1=" " ind2=" ">
    <subfield code="c">871793</subfield>
    <subfield code="a">Adaptive edge/cloud compute and network continuum over a heterogeneous sparse edge infrastructure to support nextgen applications</subfield>
  </datafield>
  <datafield tag="540" ind1=" " ind2=" ">
    <subfield code="u">https://creativecommons.org/licenses/by/4.0/legalcode</subfield>
    <subfield code="a">Creative Commons Attribution 4.0 International</subfield>
  </datafield>
  <datafield tag="650" ind1="1" ind2="7">
    <subfield code="a">cc-by</subfield>
    <subfield code="2">opendefinition.org</subfield>
  </datafield>
  <datafield tag="520" ind1=" " ind2=" ">
    <subfield code="a">&lt;p&gt;Likelihood-based generative models are a promising resource to detect out-of- distribution (OOD) inputs which could compromise the&amp;nbsp;robustness or reliability of a machine learning system. However, likelihoods derived from such models have been shown to be problematic for detecting certain types of inputs that sig- nificantly differ from training data. In this paper, we pose that this problem is due to the excessive influence that input complexity has in generative models&amp;rsquo; likelihoods. We report a set of experiments supporting this hypothesis, and use an estimate of input complexity to derive an efficient and parameter-free OOD score, which can be seen as a likelihood-ratio, akin to Bayesian model compari- son. We find such score to perform comparably to, or even better than, existing OOD detection approaches under a wide range of data sets, models, model sizes, and complexity estimates.&lt;/p&gt;</subfield>
  </datafield>
  <datafield tag="773" ind1=" " ind2=" ">
    <subfield code="n">doi</subfield>
    <subfield code="i">isVersionOf</subfield>
    <subfield code="a">10.5281/zenodo.5253739</subfield>
  </datafield>
  <datafield tag="024" ind1=" " ind2=" ">
    <subfield code="a">10.5281/zenodo.5253740</subfield>
    <subfield code="2">doi</subfield>
  </datafield>
  <datafield tag="980" ind1=" " ind2=" ">
    <subfield code="a">publication</subfield>
    <subfield code="b">conferencepaper</subfield>
  </datafield>
</record>
17
12
views
downloads
All versions This version
Views 1717
Downloads 1212
Data volume 23.3 MB23.3 MB
Unique views 1414
Unique downloads 1212

Share

Cite as