{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "lecun-denker-solla-optimal-brain-damage-1990",
    "title": "LeCun, Denker and Solla: Optimal Brain Damage",
    "summary": "A trained neural network may contain weights that contribute little to its predictions. LeCun, Denker and Solla estimate the cost of removing a weight from how sharply the training error bends near the learned solution. They remove low-cost weights and retrain. Their experiments show why curvature can be more useful than simply keeping large weights, while also showing that a local approximation becomes less accurate when too many weights are removed at once.",
    "boundary": "The NIPS contribution indexed by its publisher under 1989. The saliency estimate uses a locally quadratic loss, near-stationary parameters and a diagonal-Hessian approximation. Reported pruning and retraining results are finite experiments; the name does not assert globally optimal pruning. Network redesign and automatic weight deletion are separate sources of compression.",
    "area": "Neural and deep learning",
    "topic": "Artificial intelligence",
    "year": 1989,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "lecun-denker-solla-optimal-brain-damage-1990",
    "target": {
      "record_id": "urn:uuid:9b5ca7aa-66a0-4146-a314-3b30c78a11ca",
      "version": 1,
      "content_hash": "sha256:68c0b11d01fb412b9e138cc9b1e71209e1435f2720a0634b301292e763a0af58"
    },
    "humans": [
      "Yann LeCun",
      "John S. Denker",
      "Sara A. Solla"
    ],
    "authors_complete": true,
    "human_role_note": "Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.",
    "agents": [],
    "related": [
      "backprop-representations",
      "lecun-et-al-zip-code-recognition-1989"
    ],
    "subresults": [],
    "scientific_assessment": {
      "source_review": "SOURCE_CURATED",
      "scientific_review": "NOT_REVIEWED_BY_WRS",
      "formal_replay": "NOT_RUN",
      "provider_attestation": "NONE"
    },
    "sources": [
      {
        "id": "neurips-lecun-optimal-brain-damage-1989",
        "title": "Official NeurIPS record for Optimal Brain Damage",
        "url": "https://proceedings.neurips.cc/paper_files/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
      },
      {
        "id": "lecun-publications-optimal-brain-damage",
        "title": "Yann LeCun publication record for Optimal Brain Damage",
        "url": "https://yann.lecun.com/exdb/publis/"
      },
      {
        "id": "wrs20261023-obd-publisher-identity",
        "title": "NeurIPS publisher record for the 1989 contribution",
        "url": "https://proceedings.neurips.cc/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
      }
    ],
    "points": "240000",
    "published_grade": {
      "id": "lecun-denker-solla-optimal-brain-damage-1990",
      "entry_id": "lecun-denker-solla-optimal-brain-damage-1990",
      "band": "A",
      "modifier": "+",
      "target": {
        "record_id": "urn:uuid:9b5ca7aa-66a0-4146-a314-3b30c78a11ca",
        "version": 1,
        "content_hash": "sha256:68c0b11d01fb412b9e138cc9b1e71209e1435f2720a0634b301292e763a0af58"
      },
      "what_changed": "A trained neural network may contain weights that contribute little to its predictions. LeCun, Denker and Solla estimate the cost of removing a weight from how sharply the training error bends near the learned solution. They remove low-cost weights and retrain. Their experiments show why curvature can be more useful than simply keeping large weights, while also showing that a local approximation becomes less accurate when too many weights are removed at once.",
      "scope": "A trained neural network may contain weights that contribute little to its predictions. LeCun, Denker and Solla estimate the cost of removing a weight from how sharply the training error bends near the learned solution. They remove low-cost weights and retrain. Their experiments show why curvature can be more useful than simply keeping large weights, while also showing that a local approximation becomes less accurate when too many weights are removed at once. The NIPS contribution indexed by its publisher under 1989. The saliency estimate uses a locally quadratic loss, near-stationary parameters and a diagonal-Hessian approximation. Reported pruning and retraining results are finite experiments; the name does not assert globally optimal pruning. Network redesign and automatic weight deletion are separate sources of compression.",
      "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:9b5ca7aa-66a0-4146-a314-3b30c78a11ca",
        "version": 1,
        "content_hash": "sha256:68c0b11d01fb412b9e138cc9b1e71209e1435f2720a0634b301292e763a0af58"
      },
      "edition": "ratings-e039106f30dec177"
    }
  },
  "version": {
    "kind": "PUBLIC_RECORD_VIEW",
    "previous": null,
    "published_at": "2026-10-23T00:00:00Z",
    "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": "Yann LeCun",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Official NeurIPS record for Optimal Brain Damage",
            "url": "https://proceedings.neurips.cc/paper_files/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "John S. Denker",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Official NeurIPS record for Optimal Brain Damage",
            "url": "https://proceedings.neurips.cc/paper_files/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Sara A. Solla",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Official NeurIPS record for Optimal Brain Damage",
            "url": "https://proceedings.neurips.cc/paper_files/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "Official NeurIPS record for Optimal Brain Damage",
            "url": "https://proceedings.neurips.cc/paper_files/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "Yann LeCun publication record for Optimal Brain Damage",
            "url": "https://yann.lecun.com/exdb/publis/"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-2",
          "kind": "source_reference",
          "source": {
            "citation": "NeurIPS publisher record for the 1989 contribution",
            "url": "https://proceedings.neurips.cc/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "The NIPS contribution indexed by its publisher under 1989. The saliency estimate uses a locally quadratic loss, near-stationary parameters and a diagonal-Hessian approximation. Reported pruning and retraining results are finite experiments; the name does not assert globally optimal pruning. Network redesign and automatic weight deletion are separate sources of compression.",
        "domain": "Neural and deep learning"
      },
      "sources": [
        {
          "citation": "Official NeurIPS record for Optimal Brain Damage",
          "url": "https://proceedings.neurips.cc/paper_files/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
        },
        {
          "citation": "Yann LeCun publication record for Optimal Brain Damage",
          "url": "https://yann.lecun.com/exdb/publis/"
        },
        {
          "citation": "NeurIPS publisher record for the 1989 contribution",
          "url": "https://proceedings.neurips.cc/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html"
        }
      ],
      "statement": "# LeCun, Denker and Solla: Optimal Brain Damage\n\nA trained neural network may contain weights that contribute little to its predictions. LeCun, Denker and Solla estimate the cost of removing a weight from how sharply the training error bends near the learned solution. They remove low-cost weights and retrain. Their experiments show why curvature can be more useful than simply keeping large weights, while also showing that a local approximation becomes less accurate when too many weights are removed at once.\n\n## Scientific scope\n\nThe NIPS contribution indexed by its publisher under 1989. The saliency estimate uses a locally quadratic loss, near-stationary parameters and a diagonal-Hessian approximation. Reported pruning and retraining results are finite experiments; the name does not assert globally optimal pruning. Network redesign and automatic weight deletion are separate sources of compression.\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\nYann LeCun; John S. Denker; Sara A. Solla\n\nSource-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred.\n\n\n\n## Original sources\n\n- [Official NeurIPS record for Optimal Brain Damage](https://proceedings.neurips.cc/paper_files/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html)\n- [Yann LeCun publication record for Optimal Brain Damage](https://yann.lecun.com/exdb/publis/)\n- [NeurIPS publisher record for the 1989 contribution](https://proceedings.neurips.cc/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html)\n",
      "summary": "A trained neural network may contain weights that contribute little to its predictions. LeCun, Denker and Solla estimate the cost of removing a weight from how sharply the training error bends near the learned solution. They remove low-cost weights and retrain. Their experiments show why curvature can be more useful than simply keeping large weights, while also showing that a local approximation becomes less accurate when too many weights are removed at once.",
      "title": "LeCun, Denker and Solla: Optimal Brain Damage"
    },
    "record_id": "urn:uuid:9b5ca7aa-66a0-4146-a314-3b30c78a11ca",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:9b5ca7aa-66a0-4146-a314-3b30c78a11ca",
      "version": 1,
      "content_hash": "sha256:68c0b11d01fb412b9e138cc9b1e71209e1435f2720a0634b301292e763a0af58"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:0dca3787994aed4bdf6e9ad449d8c0aad1a0da3704d404823aa9b71172d8affc"
  },
  "target": {
    "record_id": "urn:uuid:9b5ca7aa-66a0-4146-a314-3b30c78a11ca",
    "version": 1,
    "content_hash": "sha256:68c0b11d01fb412b9e138cc9b1e71209e1435f2720a0634b301292e763a0af58"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "lecun-denker-solla-optimal-brain-damage-1990",
    "entry_id": "lecun-denker-solla-optimal-brain-damage-1990",
    "revision": 1,
    "title": "LeCun, Denker and Solla: Optimal Brain Damage",
    "context": "Neural and deep learning",
    "track": "Reusable scientific method, experimental result or historical framework",
    "band": "A",
    "modifier": "+",
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-10-23",
    "evidence_cutoff": "2026-10-23",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "A trained neural network may contain weights that contribute little to its predictions. LeCun, Denker and Solla estimate the cost of removing a weight from how sharply the training error bends near the learned solution. They remove low-cost weights and retrain. Their experiments show why curvature can be more useful than simply keeping large weights, while also showing that a local approximation becomes less accurate when too many weights are removed at once.",
    "scope": "A trained neural network may contain weights that contribute little to its predictions. LeCun, Denker and Solla estimate the cost of removing a weight from how sharply the training error bends near the learned solution. They remove low-cost weights and retrain. Their experiments show why curvature can be more useful than simply keeping large weights, while also showing that a local approximation becomes less accurate when too many weights are removed at once. The NIPS contribution indexed by its publisher under 1989. The saliency estimate uses a locally quadratic loss, near-stationary parameters and a diagonal-Hessian approximation. Reported pruning and retraining results are finite experiments; the name does not assert globally optimal pruning. Network redesign and automatic weight deletion are separate sources of compression.",
    "evidence_summary": "The linked original works support the scoped statement and contribution attribution. WRS has not independently replayed experiments or reconstructed a full mathematical proof. The NIPS contribution indexed by its publisher under 1989. The saliency estimate uses a locally quadratic loss, near-stationary parameters and a diagonal-Hessian approximation. Reported pruning and retraining results are finite experiments; the name does not assert globally optimal pruning. Network redesign and automatic weight deletion are separate sources of compression.",
    "references": [
      {
        "title": "Official NeurIPS record for Optimal Brain Damage",
        "url": "https://proceedings.neurips.cc/paper_files/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html",
        "locator": "Original scoped contribution, statement/method and evidence limits; source-specific attribution retained."
      },
      {
        "title": "Yann LeCun publication record for Optimal Brain Damage",
        "url": "https://yann.lecun.com/exdb/publis/",
        "locator": "Original scoped contribution, statement/method and evidence limits; source-specific attribution retained."
      },
      {
        "title": "NeurIPS publisher record for the 1989 contribution",
        "url": "https://proceedings.neurips.cc/paper/1989/hash/6c9882bbac1c7093bd25041881277658-Abstract.html",
        "locator": "Original scoped contribution, statement/method and evidence limits; source-specific attribution retained."
      }
    ],
    "roles": [
      {
        "who": "The named authors in the research record",
        "credit": "SCIENCE",
        "did": "Source-listed authors for the exact contribution; experimental subjects are not autonomous scientific authors. Source support and predecessor work retain separate roles. Individual effort shares and unresolved identities are not inferred."
      }
    ],
    "target": {
      "record_id": "urn:uuid:9b5ca7aa-66a0-4146-a314-3b30c78a11ca",
      "version": 1,
      "content_hash": "sha256:68c0b11d01fb412b9e138cc9b1e71209e1435f2720a0634b301292e763a0af58"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "240000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}