{ "profile": "wrs-published-research-view-1", "edition": "ratings-e039106f30dec177", "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS", "entry": { "id": "ai-oct-neural-operators-gkn-fno", "title": "Graph-kernel and Fourier neural operators", "summary": "Learns parameterized maps between input and solution functions through shared integral-operator layers; GKN approximates kernels by graph message passing and Nyström sampling, while FNO parameterizes truncated Fourier modes for efficient grid evaluation.", "boundary": "Resolution-transfer and solver-speed results are empirical and exclude training/data-generation cost; FNO FFT implementation assumes a uniform grid and truncated periodic kernel. No universal trained-solver accuracy or existence-of-optimal-training guarantee.", "area": "Neural operators and scientific machine learning", "topic": "Artificial intelligence", "year": 2020, "lane": "HUMAN", "result": "ALGORITHM", "package_id": "ai-oct-neural-operators-gkn-fno", "target": { "record_id": "urn:uuid:df1a5b95-1137-47bd-ab51-b8acd2ef67f9", "version": 1, "content_hash": "sha256:1e21006a9e458e46d492d7bbb23b2ca24ec27c2af7cd4e6fa942b577693f30cc" }, "humans": [ "Zongyi Li", "Nikola Kovachki", "Kamyar Azizzadenesheli", "Burigede Liu", "Kaushik Bhattacharya", "Andrew Stuart", "Anima Anandkumar" ], "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-2003-03485v1-primary", "title": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" }, { "id": "src-ai-oct-2010-08895v1-primary", "title": "Fourier Neural Operator for Parametric Partial Differential Equations", "url": "https://arxiv.org/pdf/2010.08895v1" } ], "points": "320000", "published_grade": { "id": "ai-oct-neural-operators-gkn-fno", "entry_id": "ai-oct-neural-operators-gkn-fno", "band": "S", "modifier": "-", "target": { "record_id": "urn:uuid:df1a5b95-1137-47bd-ab51-b8acd2ef67f9", "version": 1, "content_hash": "sha256:1e21006a9e458e46d492d7bbb23b2ca24ec27c2af7cd4e6fa942b577693f30cc" }, "what_changed": "Learns parameterized maps between input and solution functions through shared integral-operator layers; GKN approximates kernels by graph message passing and Nyström sampling, while FNO parameterizes truncated Fourier modes for efficient grid evaluation.", "scope": "Learns parameterized maps between input and solution functions through shared integral-operator layers; GKN approximates kernels by graph message passing and Nyström sampling, while FNO parameterizes truncated Fourier modes for efficient grid evaluation. Resolution-transfer and solver-speed results are empirical and exclude training/data-generation cost; FNO FFT implementation assumes a uniform grid and truncated periodic kernel. No universal trained-solver accuracy or existence-of-optimal-training 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:df1a5b95-1137-47bd-ab51-b8acd2ef67f9", "version": 1, "content_hash": "sha256:1e21006a9e458e46d492d7bbb23b2ca24ec27c2af7cd4e6fa942b577693f30cc" }, "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": "Zongyi Li", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Nikola Kovachki", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Kamyar Azizzadenesheli", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Burigede Liu", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Kaushik Bhattacharya", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Andrew Stuart", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Anima Anandkumar", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" } } ], "evidence": [ { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-0", "kind": "source_reference", "source": { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" } }, { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-1", "kind": "source_reference", "source": { "citation": "Fourier Neural Operator for Parametric Partial Differential Equations", "url": "https://arxiv.org/pdf/2010.08895v1" } } ], "origin": "HUMAN_LEGACY", "record_profile": "wrs-local-research-notes-0.1", "record_type": "research_note", "rrp_version": "0.1", "scope": { "description": "Resolution-transfer and solver-speed results are empirical and exclude training/data-generation cost; FNO FFT implementation assumes a uniform grid and truncated periodic kernel. No universal trained-solver accuracy or existence-of-optimal-training guarantee.", "domain": "Neural operators and scientific machine learning" }, "sources": [ { "citation": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1" }, { "citation": "Fourier Neural Operator for Parametric Partial Differential Equations", "url": "https://arxiv.org/pdf/2010.08895v1" } ], "statement": "# Graph-kernel and Fourier neural operators\n\n## The question\n\nAn ordinary prediction model might map a fixed list of inputs to a fixed list of outputs. A physical problem often asks for something larger: take an entire initial field or material-coefficient function and predict an entire solution field. Neural operators aim to learn this function-to-function relationship, so the idea is not tied only to one numbered list of grid points.\n\n## The idea\n\nThe source family develops shared operator layers, including graph-based kernel constructions and Fourier-domain layers. Fourier methods express patterns as combinations of waves. A learned layer can transform selected frequency components and combine them with local processing. Sharing this structure across discretizations creates a route to studying resolution transfer, rather than training unrelated models for every grid size.\n\n## Why it matters\n\nThe important advance is a reusable approach to learning families of physical maps. It is separate from proving that a particular trained network accurately solves every equation. Training examples usually come from observations or numerical solvers, and producing them costs resources. A rapid forward prediction is not the full cost of building the model or establishing where it can be trusted.\n\n## A teaching example\n\nFor an original classroom picture, imagine predicting the temperature across a metal plate after an initial heat pattern is supplied. A coarse grid and a fine grid describe the same underlying field at different sampling points. Transferring a learned rule between them is attractive, but fine-scale patterns absent from training may still be missed. Fourier truncation deliberately limits which frequency information is represented.\n\n## Scope and a useful next step\n\nChoose one small published equation example and separate training, interpolation and out-of-distribution tests. Record grid spacing, boundary conditions and reference-solver error. Compare predictions at a second resolution without hiding the data-generation cost. A finite successful test supports that stated range; it does not certify arbitrary-resolution accuracy or a universal replacement for numerical analysis. Resolution-transfer and solver-speed results are empirical and exclude training/data-generation cost; FNO FFT implementation assumes a uniform grid and truncated periodic kernel. No universal trained-solver accuracy or existence-of-optimal-training guarantee.\n\n## Scientific scope\n\nResolution-transfer and solver-speed results are empirical and exclude training/data-generation cost; FNO FFT implementation assumes a uniform grid and truncated periodic kernel. No universal trained-solver accuracy or existence-of-optimal-training guarantee.\n\n## Sources for this explanation\n\n- [Neural Operator: Graph Kernel Network for Partial Differential Equations](https://arxiv.org/pdf/2003.03485v1): Pinned original contribution and its stated method, assumptions and limits. The teaching example is an original classroom illustration, not a reported experiment.\n- [Fourier Neural Operator for Parametric Partial Differential Equations](https://arxiv.org/pdf/2010.08895v1): Pinned original contribution and its stated method, assumptions and limits. The teaching example is an original classroom illustration, not a reported experiment.\n", "summary": "Learns parameterized maps between input and solution functions through shared integral-operator layers; GKN approximates kernels by graph message passing and Nyström sampling, while FNO parameterizes truncated Fourier modes for efficient grid evaluation.", "title": "Graph-kernel and Fourier neural operators" }, "record_id": "urn:uuid:df1a5b95-1137-47bd-ab51-b8acd2ef67f9", "version": 1, "source_ref": { "record_id": "urn:uuid:df1a5b95-1137-47bd-ab51-b8acd2ef67f9", "version": 1, "content_hash": "sha256:1e21006a9e458e46d492d7bbb23b2ca24ec27c2af7cd4e6fa942b577693f30cc" }, "export_view": "wrs-public-editorial-view-1", "public_view_hash": "sha256:e35ff5e9c9887b8637be036dfa5a0d799c2e2a6d4d5172333290444d2d535f81" }, "target": { "record_id": "urn:uuid:df1a5b95-1137-47bd-ab51-b8acd2ef67f9", "version": 1, "content_hash": "sha256:1e21006a9e458e46d492d7bbb23b2ca24ec27c2af7cd4e6fa942b577693f30cc" }, "published_grade": { "profile": "wrs-published-grade-1", "edition": "2026-10-28.1", "id": "ai-oct-neural-operators-gkn-fno", "entry_id": "ai-oct-neural-operators-gkn-fno", "revision": 1, "title": "Graph-kernel and Fourier neural operators", "context": "Neural operators and scientific machine learning", "track": "Human-designed research method or resource", "band": "S", "modifier": "-", "status": "PROVISIONAL_EDITORIAL", "assessed_on": "2026-10-16", "evidence_cutoff": "2026-10-16", "achievement_state": "SUPPORTED_SCOPED_RESULT", "what_changed": "Learns parameterized maps between input and solution functions through shared integral-operator layers; GKN approximates kernels by graph message passing and Nyström sampling, while FNO parameterizes truncated Fourier modes for efficient grid evaluation.", "scope": "Learns parameterized maps between input and solution functions through shared integral-operator layers; GKN approximates kernels by graph message passing and Nyström sampling, while FNO parameterizes truncated Fourier modes for efficient grid evaluation. Resolution-transfer and solver-speed results are empirical and exclude training/data-generation cost; FNO FFT implementation assumes a uniform grid and truncated periodic kernel. No universal trained-solver accuracy or existence-of-optimal-training guarantee.", "evidence_summary": "The linked original sources describe this contribution and its evidence limits. WRS has not independently reconstructed the full scientific claims. Resolution-transfer and solver-speed results are empirical and exclude training/data-generation cost; FNO FFT implementation assumes a uniform grid and truncated periodic kernel. No universal trained-solver accuracy or existence-of-optimal-training guarantee.", "references": [ { "title": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1", "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification." }, { "title": "Fourier Neural Operator for Parametric Partial Differential Equations", "url": "https://arxiv.org/pdf/2010.08895v1", "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:df1a5b95-1137-47bd-ab51-b8acd2ef67f9", "version": 1, "content_hash": "sha256:1e21006a9e458e46d492d7bbb23b2ca24ec27c2af7cd4e6fa942b577693f30cc" }, "scientific_correctness_certified": false, "independent_expert_calibration": false, "method": "PROPRIETARY", "points": "320000" }, "canonical_agent_submission": false, "scientific_verification_awarded": false, "reader_guide": { "id": "ai-oct-neural-operators-gkn-fno", "target": { "record_id": "urn:uuid:df1a5b95-1137-47bd-ab51-b8acd2ef67f9", "version": 1, "content_hash": "sha256:1e21006a9e458e46d492d7bbb23b2ca24ec27c2af7cd4e6fa942b577693f30cc" }, "sections": [ { "title": "The question", "text": "An ordinary prediction model might map a fixed list of inputs to a fixed list of outputs. A physical problem often asks for something larger: take an entire initial field or material-coefficient function and predict an entire solution field. Neural operators aim to learn this function-to-function relationship, so the idea is not tied only to one numbered list of grid points." }, { "title": "The idea", "text": "The source family develops shared operator layers, including graph-based kernel constructions and Fourier-domain layers. Fourier methods express patterns as combinations of waves. A learned layer can transform selected frequency components and combine them with local processing. Sharing this structure across discretizations creates a route to studying resolution transfer, rather than training unrelated models for every grid size." }, { "title": "Why it matters", "text": "The important advance is a reusable approach to learning families of physical maps. It is separate from proving that a particular trained network accurately solves every equation. Training examples usually come from observations or numerical solvers, and producing them costs resources. A rapid forward prediction is not the full cost of building the model or establishing where it can be trusted." }, { "title": "A teaching example", "text": "For an original classroom picture, imagine predicting the temperature across a metal plate after an initial heat pattern is supplied. A coarse grid and a fine grid describe the same underlying field at different sampling points. Transferring a learned rule between them is attractive, but fine-scale patterns absent from training may still be missed. Fourier truncation deliberately limits which frequency information is represented." }, { "title": "Scope and a useful next step", "text": "Choose one small published equation example and separate training, interpolation and out-of-distribution tests. Record grid spacing, boundary conditions and reference-solver error. Compare predictions at a second resolution without hiding the data-generation cost. A finite successful test supports that stated range; it does not certify arbitrary-resolution accuracy or a universal replacement for numerical analysis. Resolution-transfer and solver-speed results are empirical and exclude training/data-generation cost; FNO FFT implementation assumes a uniform grid and truncated periodic kernel. No universal trained-solver accuracy or existence-of-optimal-training guarantee." } ], "references": [ { "title": "Neural Operator: Graph Kernel Network for Partial Differential Equations", "url": "https://arxiv.org/pdf/2003.03485v1", "locator": "Pinned original contribution and its stated method, assumptions and limits. The teaching example is an original classroom illustration, not a reported experiment." }, { "title": "Fourier Neural Operator for Parametric Partial Differential Equations", "url": "https://arxiv.org/pdf/2010.08895v1", "locator": "Pinned original contribution and its stated method, assumptions and limits. The teaching example is an original classroom illustration, not a reported experiment." } ] }, "reading_status": "EXPLANATORY_GUIDE" }