{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "ai-oct-mc-dropout",
    "title": "MC dropout: variational uncertainty from stochastic forward passes",
    "summary": "Interprets dropout networks through an approximate variational Bayesian model and estimates predictive moments by retaining random dropout masks during repeated inference passes.",
    "boundary": "Not invention of dropout or model averaging, nor exact Bayesian posterior inference. The approximation depends on model, likelihood, regularization and dropout choices; T forward passes still incur T-sample compute despite possible parallel latency. No universal uncertainty-calibration guarantee.",
    "area": "AI uncertainty, reliability and selective prediction",
    "topic": "Artificial intelligence",
    "year": 2015,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "ai-oct-mc-dropout",
    "target": {
      "record_id": "urn:uuid:c506b431-3d18-4b91-a3ff-dbe0b739a47e",
      "version": 1,
      "content_hash": "sha256:4e066923994a0edc5de6e21936339fcc631e617903ebf9e14c1794f5e890577e"
    },
    "humans": [
      "Yarin Gal",
      "Zoubin Ghahramani"
    ],
    "authors_complete": true,
    "human_role_note": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.",
    "agents": [],
    "related": [],
    "subresults": [],
    "scientific_assessment": {
      "source_review": "SOURCE_CURATED",
      "scientific_review": "NOT_REVIEWED_BY_WRS",
      "formal_replay": "NOT_RUN",
      "provider_attestation": "NONE"
    },
    "sources": [
      {
        "id": "src-ai-oct-mc-dropout-1506-02142v1",
        "title": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
        "url": "https://arxiv.org/pdf/1506.02142v1"
      }
    ],
    "points": "150000",
    "published_grade": {
      "id": "ai-oct-mc-dropout",
      "entry_id": "ai-oct-mc-dropout",
      "band": "A",
      "modifier": null,
      "target": {
        "record_id": "urn:uuid:c506b431-3d18-4b91-a3ff-dbe0b739a47e",
        "version": 1,
        "content_hash": "sha256:4e066923994a0edc5de6e21936339fcc631e617903ebf9e14c1794f5e890577e"
      },
      "what_changed": "Interprets dropout networks through an approximate variational Bayesian model and estimates predictive moments by retaining random dropout masks during repeated inference passes.",
      "scope": "Interprets dropout networks through an approximate variational Bayesian model and estimates predictive moments by retaining random dropout masks during repeated inference passes. Not invention of dropout or model averaging, nor exact Bayesian posterior inference. The approximation depends on model, likelihood, regularization and dropout choices; T forward passes still incur T-sample compute despite possible parallel latency. No universal uncertainty-calibration guarantee.",
      "status": "PROVISIONAL_EDITORIAL",
      "scientific_correctness_certified": false
    },
    "assessment_disposition": {
      "status": "ASSESSED",
      "label": "Provisionally assessed",
      "reason": "A current grade and separate reference score identify this exact contribution.",
      "target": {
        "record_id": "urn:uuid:c506b431-3d18-4b91-a3ff-dbe0b739a47e",
        "version": 1,
        "content_hash": "sha256:4e066923994a0edc5de6e21936339fcc631e617903ebf9e14c1794f5e890577e"
      },
      "edition": "ratings-e039106f30dec177"
    }
  },
  "version": {
    "kind": "PUBLIC_RECORD_VIEW",
    "previous": null,
    "published_at": "2026-10-16T00:00:00Z",
    "publisher": {
      "principal_id": "urn:uuid:80116931-785b-40a2-a24b-9d59db0c7e33",
      "role": "WRS_EDITORIAL_PUBLICATION"
    },
    "record": {
      "contributors": [
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Yarin Gal",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
            "url": "https://arxiv.org/pdf/1506.02142v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Zoubin Ghahramani",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
            "url": "https://arxiv.org/pdf/1506.02142v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
            "url": "https://arxiv.org/pdf/1506.02142v1"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "Not invention of dropout or model averaging, nor exact Bayesian posterior inference. The approximation depends on model, likelihood, regularization and dropout choices; T forward passes still incur T-sample compute despite possible parallel latency. No universal uncertainty-calibration guarantee.",
        "domain": "AI uncertainty, reliability and selective prediction"
      },
      "sources": [
        {
          "citation": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
          "url": "https://arxiv.org/pdf/1506.02142v1"
        }
      ],
      "statement": "# MC dropout: variational uncertainty from stochastic forward passes\n\nInterprets dropout networks through an approximate variational Bayesian model and estimates predictive moments by retaining random dropout masks during repeated inference passes.\n\n## Scientific scope\n\nNot invention of dropout or model averaging, nor exact Bayesian posterior inference. The approximation depends on model, likelihood, regularization and dropout choices; T forward passes still incur T-sample compute despite possible parallel latency. No universal uncertainty-calibration guarantee.\n\n## Reading status\n\nThis is a source-linked WRS catalogue note. A full explanatory guide and independent scientific reconstruction have not been completed for this entry. Read the identified original work for its complete narrative and arguments.\n\n## People and systems\n\nYarin Gal; Zoubin Ghahramani\n\nSource-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.\n\n\n\n## Original sources\n\n- [Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning](https://arxiv.org/pdf/1506.02142v1)\n",
      "summary": "Interprets dropout networks through an approximate variational Bayesian model and estimates predictive moments by retaining random dropout masks during repeated inference passes.",
      "title": "MC dropout: variational uncertainty from stochastic forward passes"
    },
    "record_id": "urn:uuid:c506b431-3d18-4b91-a3ff-dbe0b739a47e",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:c506b431-3d18-4b91-a3ff-dbe0b739a47e",
      "version": 1,
      "content_hash": "sha256:4e066923994a0edc5de6e21936339fcc631e617903ebf9e14c1794f5e890577e"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:c88fce1868077cd1bd24f48415f6d352a0b6d69c34a748f4b20651a3753d425f"
  },
  "target": {
    "record_id": "urn:uuid:c506b431-3d18-4b91-a3ff-dbe0b739a47e",
    "version": 1,
    "content_hash": "sha256:4e066923994a0edc5de6e21936339fcc631e617903ebf9e14c1794f5e890577e"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "ai-oct-mc-dropout",
    "entry_id": "ai-oct-mc-dropout",
    "revision": 1,
    "title": "MC dropout: variational uncertainty from stochastic forward passes",
    "context": "AI uncertainty, reliability and selective prediction",
    "track": "Human-designed research method or resource",
    "band": "A",
    "modifier": null,
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-10-16",
    "evidence_cutoff": "2026-10-16",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Interprets dropout networks through an approximate variational Bayesian model and estimates predictive moments by retaining random dropout masks during repeated inference passes.",
    "scope": "Interprets dropout networks through an approximate variational Bayesian model and estimates predictive moments by retaining random dropout masks during repeated inference passes. Not invention of dropout or model averaging, nor exact Bayesian posterior inference. The approximation depends on model, likelihood, regularization and dropout choices; T forward passes still incur T-sample compute despite possible parallel latency. No universal uncertainty-calibration guarantee.",
    "evidence_summary": "The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Not invention of dropout or model averaging, nor exact Bayesian posterior inference. The approximation depends on model, likelihood, regularization and dropout choices; T forward passes still incur T-sample compute despite possible parallel latency. No universal uncertainty-calibration guarantee.",
    "references": [
      {
        "title": "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning",
        "url": "https://arxiv.org/pdf/1506.02142v1",
        "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification."
      }
    ],
    "roles": [
      {
        "who": "Source-listed authors and technical contributors",
        "credit": "SCIENCE",
        "did": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities."
      }
    ],
    "target": {
      "record_id": "urn:uuid:c506b431-3d18-4b91-a3ff-dbe0b739a47e",
      "version": 1,
      "content_hash": "sha256:4e066923994a0edc5de6e21936339fcc631e617903ebf9e14c1794f5e890577e"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "150000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}