Report Open Access

Fail-Safe Execution of Deep Learning based Systems through Uncertainty Monitoring

Michael Weiss; Paolo Tonella


JSON Export

{
  "files": [
    {
      "links": {
        "self": "https://zenodo.org/api/files/b888d9ff-c3c1-432b-abc2-49b0218a7c5c/TR-Precrime-2020-05.pdf"
      }, 
      "checksum": "md5:ed486374270224a43e2ebc6cf3cd2a5e", 
      "bucket": "b888d9ff-c3c1-432b-abc2-49b0218a7c5c", 
      "key": "TR-Precrime-2020-05.pdf", 
      "type": "pdf", 
      "size": 903143
    }
  ], 
  "owners": [
    239303
  ], 
  "doi": "10.5281/zenodo.5055710", 
  "stats": {
    "version_unique_downloads": 25.0, 
    "unique_views": 23.0, 
    "views": 27.0, 
    "version_views": 27.0, 
    "unique_downloads": 25.0, 
    "version_unique_views": 23.0, 
    "volume": 25288004.0, 
    "version_downloads": 28.0, 
    "downloads": 28.0, 
    "version_volume": 25288004.0
  }, 
  "links": {
    "doi": "https://doi.org/10.5281/zenodo.5055710", 
    "conceptdoi": "https://doi.org/10.5281/zenodo.5055709", 
    "bucket": "https://zenodo.org/api/files/b888d9ff-c3c1-432b-abc2-49b0218a7c5c", 
    "conceptbadge": "https://zenodo.org/badge/doi/10.5281/zenodo.5055709.svg", 
    "html": "https://zenodo.org/record/5055710", 
    "latest_html": "https://zenodo.org/record/5055710", 
    "badge": "https://zenodo.org/badge/doi/10.5281/zenodo.5055710.svg", 
    "latest": "https://zenodo.org/api/records/5055710"
  }, 
  "conceptdoi": "10.5281/zenodo.5055709", 
  "created": "2021-07-01T15:29:44.813853+00:00", 
  "updated": "2021-07-02T01:48:17.897649+00:00", 
  "conceptrecid": "5055709", 
  "revision": 2, 
  "id": 5055710, 
  "metadata": {
    "access_right_category": "success", 
    "doi": "10.5281/zenodo.5055710", 
    "description": "<p>Modern software systems rely on Deep Neural Networks (DNN) when processing complex, unstructured inputs, such as images, videos, natural language texts or audio signals.&nbsp;Provided the intractably large size of such input spaces, the intrinsic limitations of learning algorithms&nbsp; and the ambiguity about the expected predictions for some of the inputs, not only there is no guarantee that DNN&#39;s predictions are always correct, but rather developers must safely assume a low, though not negligible, error probability.&nbsp;A fail-safe Deep Learning based System (DLS) is one equipped to handle DNN faults by means of a supervisor, capable of recognizing predictions that should not be trusted and that should activate a healing procedure bringing the DLS to a safe state.</p>\n\n<p>In this paper, we propose an approach to use DNN uncertainty estimators to implement such supervisor.&nbsp;We first discuss advantages and disadvantages of existing approaches to measure uncertainty for DNNs&nbsp;and propose novel metrics for the empirical assessment of the&nbsp; supervisor that rely on such approaches.&nbsp;We then describe our publicly available tool Uncertainty-Wizard, which allows transparent estimation of uncertainty for regular tf.keras DNNs.&nbsp;Lastly, we discuss a large-scale&nbsp; study conducted on four different subjects to empirically validate the approach,&nbsp;reporting the lessons-learned as guidance for software engineers who intend to monitor uncertainty for fail-safe execution of DLS.</p>", 
    "license": {
      "id": "CC-BY-4.0"
    }, 
    "title": "Fail-Safe Execution of Deep Learning based Systems through Uncertainty Monitoring", 
    "relations": {
      "version": [
        {
          "count": 1, 
          "index": 0, 
          "parent": {
            "pid_type": "recid", 
            "pid_value": "5055709"
          }, 
          "is_last": true, 
          "last_child": {
            "pid_type": "recid", 
            "pid_value": "5055710"
          }
        }
      ]
    }, 
    "grants": [
      {
        "code": "787703", 
        "links": {
          "self": "https://zenodo.org/api/grants/10.13039/501100000780::787703"
        }, 
        "title": "Self-assessment Oracles for Anticipatory Testing", 
        "acronym": "PRECRIME", 
        "program": "H2020", 
        "funder": {
          "doi": "10.13039/501100000780", 
          "acronyms": [], 
          "name": "European Commission", 
          "links": {
            "self": "https://zenodo.org/api/funders/10.13039/501100000780"
          }
        }
      }
    ], 
    "publication_date": "2020-09-01", 
    "creators": [
      {
        "affiliation": "Universit\u00e0 della Svizzera italiana", 
        "name": "Michael Weiss"
      }, 
      {
        "affiliation": "Universit\u00e0 della Svizzera italiana", 
        "name": "Paolo Tonella"
      }
    ], 
    "access_right": "open", 
    "resource_type": {
      "subtype": "report", 
      "type": "publication", 
      "title": "Report"
    }, 
    "related_identifiers": [
      {
        "scheme": "doi", 
        "identifier": "10.1109/ICST49551.2021.00015", 
        "relation": "isObsoletedBy"
      }, 
      {
        "scheme": "doi", 
        "identifier": "10.5281/zenodo.5055709", 
        "relation": "isVersionOf"
      }
    ]
  }
}
27
28
views
downloads
All versions This version
Views 2727
Downloads 2828
Data volume 25.3 MB25.3 MB
Unique views 2323
Unique downloads 2525

Share

Cite as