{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:YBLT2AS6PZ4HFYMQ2BDDXOLPGX","short_pith_number":"pith:YBLT2AS6","schema_version":"1.0","canonical_sha256":"c0573d025e7e7872e190d0463bb96f35e7152ffe5d780dea8df79f8fd9b825dc","source":{"kind":"arxiv","id":"2405.15603","version":3},"attestation_state":"computed","paper":{"title":"Kronecker-Factored Approximate Curvature for Physics-Informed Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["physics.comp-ph"],"primary_cat":"cs.LG","authors_text":"Felix Dangel, Johannes M\\\"uller, Marius Zeinhofer","submitted_at":"2024-05-24T14:36:02Z","abstract_excerpt":"Physics-informed neural networks (PINNs) are infamous for being hard to train. Recently, second-order methods based on natural gradient and Gauss-Newton methods have shown promising performance, improving the accuracy achieved by first-order methods by several orders of magnitude. While promising, the proposed methods only scale to networks with a few thousand parameters due to the high computational cost to evaluate, store, and invert the curvature matrix. We propose Kronecker-factored approximate curvature (KFAC) for PINN losses that greatly reduces the computational cost and allows scaling "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2405.15603","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-24T14:36:02Z","cross_cats_sorted":["physics.comp-ph"],"title_canon_sha256":"22c4df3d0647639fc88d7cc3aa125ad2793bf68286b9cd56be056f12d3ffa803","abstract_canon_sha256":"af82cc9f08d086b4959053916444ef90b0244f6b7a130c80a31fc5c821297333"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:28:31.196270Z","signature_b64":"BXA8Av83ll3WUwwIPkvjZImeazuZFWHAzrYbwmgXqiWFuEvLtg4HG2RzJm7UsdKkMJAm9/COYqAegk//0Q6bAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c0573d025e7e7872e190d0463bb96f35e7152ffe5d780dea8df79f8fd9b825dc","last_reissued_at":"2026-07-05T09:28:31.195831Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:28:31.195831Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Kronecker-Factored Approximate Curvature for Physics-Informed Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["physics.comp-ph"],"primary_cat":"cs.LG","authors_text":"Felix Dangel, Johannes M\\\"uller, Marius Zeinhofer","submitted_at":"2024-05-24T14:36:02Z","abstract_excerpt":"Physics-informed neural networks (PINNs) are infamous for being hard to train. Recently, second-order methods based on natural gradient and Gauss-Newton methods have shown promising performance, improving the accuracy achieved by first-order methods by several orders of magnitude. While promising, the proposed methods only scale to networks with a few thousand parameters due to the high computational cost to evaluate, store, and invert the curvature matrix. We propose Kronecker-factored approximate curvature (KFAC) for PINN losses that greatly reduces the computational cost and allows scaling "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.15603","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2405.15603/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2405.15603","created_at":"2026-07-05T09:28:31.195880+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.15603v3","created_at":"2026-07-05T09:28:31.195880+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.15603","created_at":"2026-07-05T09:28:31.195880+00:00"},{"alias_kind":"pith_short_12","alias_value":"YBLT2AS6PZ4H","created_at":"2026-07-05T09:28:31.195880+00:00"},{"alias_kind":"pith_short_16","alias_value":"YBLT2AS6PZ4HFYMQ","created_at":"2026-07-05T09:28:31.195880+00:00"},{"alias_kind":"pith_short_8","alias_value":"YBLT2AS6","created_at":"2026-07-05T09:28:31.195880+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.02194","citing_title":"An Optimisation Framework for the Well-Conditioned Training of Physics-Informed Neural Networks","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX","json":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX.json","graph_json":"https://pith.science/api/pith-number/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/graph.json","events_json":"https://pith.science/api/pith-number/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/events.json","paper":"https://pith.science/paper/YBLT2AS6"},"agent_actions":{"view_html":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX","download_json":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX.json","view_paper":"https://pith.science/paper/YBLT2AS6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.15603&json=true","fetch_graph":"https://pith.science/api/pith-number/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/graph.json","fetch_events":"https://pith.science/api/pith-number/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/action/storage_attestation","attest_author":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/action/author_attestation","sign_citation":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/action/citation_signature","submit_replication":"https://pith.science/pith/YBLT2AS6PZ4HFYMQ2BDDXOLPGX/action/replication_record"}},"created_at":"2026-07-05T09:28:31.195880+00:00","updated_at":"2026-07-05T09:28:31.195880+00:00"}