Operationally meaningful representations of physical systems in neural networks

Poulsen Nautrup, Hendrik; Metger, Tony; Iten, Raban; Jerbi, Sofiene; Trenkwalder, Lea M.; Wilming, Henrik; Briegel, Hans J.; Renner, Renato

Operationally meaningful representations of physical systems in neural networks

Files

Poulsen Nautrup_2-uzosr6ugoajk8.PDF(2.48 MB)

Date

2022

Authors

Poulsen Nautrup, Hendrik

Metger, Tony

Iten, Raban

Jerbi, Sofiene

Trenkwalder, Lea M.

Wilming, Henrik

Briegel, Hans J.

Renner, Renato

Publication type

Journal article

Publication status

Published

Published in

Machine Learning: Science and Technology ; 3 (2022), 4. - 045025. - IOP Publishing. - eISSN 2632-2153

Abstract

To make progress in science, we often build abstract representations of physical systems that meaningfully encode information about the systems. Such representations ignore redundant features and treat parameters such as velocity and position separately because they can be useful for making statements about different experimental settings. Here, we capture this notion by formally defining the concept of operationally meaningful representations. We present an autoencoder architecture with attention mechanism that can generate such representations and demonstrate it on examples involving both classical and quantum physics. For instance, our architecture finds a compact representation of an arbitrary two-qubit system that separates local parameters from parameters describing quantum correlations.

Subject (DDC)

100 Philosophy

Keywords

representation learning, neural networks, reinforcement learning, Bloch vector, quantum physics

Cite This

ISO 690

POULSEN NAUTRUP, Hendrik, Tony METGER, Raban ITEN, Sofiene JERBI, Lea M. TRENKWALDER, Henrik WILMING, Hans J. BRIEGEL, Renato RENNER, 2022. Operationally meaningful representations of physical systems in neural networks. In: Machine Learning: Science and Technology. IOP Publishing. 3(4), 045025. eISSN 2632-2153. Available under: doi: 10.1088/2632-2153/ac9ae8

BibTex

@article{PoulsenNautrup2022Opera-59643,
  year={2022},
  doi={10.1088/2632-2153/ac9ae8},
  title={Operationally meaningful representations of physical systems in neural networks},
  number={4},
  volume={3},
  journal={Machine Learning: Science and Technology},
  author={Poulsen Nautrup, Hendrik and Metger, Tony and Iten, Raban and Jerbi, Sofiene and Trenkwalder, Lea M. and Wilming, Henrik and Briegel, Hans J. and Renner, Renato},
  note={Article Number: 045025}
}

RDF

<rdf:RDF
    xmlns:dcterms="http://purl.org/dc/terms/"
    xmlns:dc="http://purl.org/dc/elements/1.1/"
    xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
    xmlns:bibo="http://purl.org/ontology/bibo/"
    xmlns:dspace="http://digital-repositories.org/ontologies/dspace/0.1.0#"
    xmlns:foaf="http://xmlns.com/foaf/0.1/"
    xmlns:void="http://rdfs.org/ns/void#"
    xmlns:xsd="http://www.w3.org/2001/XMLSchema#" > 
  <rdf:Description rdf:about="https://kops.uni-konstanz.de/server/rdf/resource/123456789/59643">
    <dc:creator>Renner, Renato</dc:creator>
    <dc:language>eng</dc:language>
    <dc:contributor>Iten, Raban</dc:contributor>
    <dcterms:hasPart rdf:resource="https://kops.uni-konstanz.de/bitstream/123456789/59643/1/Poulsen%20Nautrup_2-uzosr6ugoajk8.PDF"/>
    <dcterms:issued>2022</dcterms:issued>
    <dc:creator>Trenkwalder, Lea M.</dc:creator>
    <dc:creator>Jerbi, Sofiene</dc:creator>
    <dc:creator>Briegel, Hans J.</dc:creator>
    <dc:creator>Metger, Tony</dc:creator>
    <void:sparqlEndpoint rdf:resource="http://localhost/fuseki/dspace/sparql"/>
    <foaf:homepage rdf:resource="http://localhost:8080/"/>
    <dc:contributor>Poulsen Nautrup, Hendrik</dc:contributor>
    <dc:creator>Iten, Raban</dc:creator>
    <dc:creator>Wilming, Henrik</dc:creator>
    <dc:contributor>Renner, Renato</dc:contributor>
    <dc:contributor>Briegel, Hans J.</dc:contributor>
    <dc:contributor>Jerbi, Sofiene</dc:contributor>
    <dc:contributor>Wilming, Henrik</dc:contributor>
    <dc:date rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2023-01-05T11:23:23Z</dc:date>
    <dspace:isPartOfCollection rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/40"/>
    <dc:contributor>Metger, Tony</dc:contributor>
    <dc:creator>Poulsen Nautrup, Hendrik</dc:creator>
    <dcterms:available rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2023-01-05T11:23:23Z</dcterms:available>
    <bibo:uri rdf:resource="https://kops.uni-konstanz.de/handle/123456789/59643"/>
    <dcterms:abstract xml:lang="eng">To make progress in science, we often build abstract representations of physical systems that meaningfully encode information about the systems. Such representations ignore redundant features and treat parameters such as velocity and position separately because they can be useful for making statements about different experimental settings. Here, we capture this notion by formally defining the concept of operationally meaningful representations. We present an autoencoder architecture with attention mechanism that can generate such representations and demonstrate it on examples involving both classical and quantum physics. For instance, our architecture finds a compact representation of an arbitrary two-qubit system that separates local parameters from parameters describing quantum correlations.</dcterms:abstract>
    <dcterms:title>Operationally meaningful representations of physical systems in neural networks</dcterms:title>
    <dc:rights>terms-of-use</dc:rights>
    <dspace:hasBitstream rdf:resource="https://kops.uni-konstanz.de/bitstream/123456789/59643/1/Poulsen%20Nautrup_2-uzosr6ugoajk8.PDF"/>
    <dc:contributor>Trenkwalder, Lea M.</dc:contributor>
    <dcterms:isPartOf rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/40"/>
    <dcterms:rights rdf:resource="https://rightsstatements.org/page/InC/1.0/"/>
  </rdf:Description>
</rdf:RDF>

Refereed

Yes