{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:3QF254ZSZFGISYXIPDO3UHM6SC","short_pith_number":"pith:3QF254ZS","schema_version":"1.0","canonical_sha256":"dc0baef332c94c8962e878ddba1d9e9092a542d23b5617dd5c8e9fa321da36d0","source":{"kind":"arxiv","id":"2408.00573","version":4},"attestation_state":"computed","paper":{"title":"Convergence Analysis of Natural Gradient Descent for Over-parameterized Physics-Informed Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bin Shan, Ting Du, Wang Kong, Xianliang Xu, Ye Li, Zhongyi Huang","submitted_at":"2024-08-01T14:06:34Z","abstract_excerpt":"In the context of over-parameterization, there is a line of work demonstrating that randomly initialized (stochastic) gradient descent (GD) converges to a globally optimal solution at a linear convergence rate for the quadratic loss function. However, the learning rate of GD for training two-layer neural networks exhibits poor dependence on the sample size and the Gram matrix, leading to a slow training process. In this paper, we show that for training two-layer $\\text{ReLU}^3$ Physics-Informed Neural Networks (PINNs), the learning rate can be improved from $\\mathcal{O}(\\lambda_0)$ to $\\mathca"},"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":"2408.00573","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-08-01T14:06:34Z","cross_cats_sorted":[],"title_canon_sha256":"38f40f0b4be2c49c47f5738e29a8d9e821d4a6675e23f0d6c8a462055f272f2e","abstract_canon_sha256":"389c1755ef4f641f4fc78db20876acecdbc9d6c04d4534166712fe4566b060d6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:45.647926Z","signature_b64":"5qZ4KeFkhElVw5i/eZu9wenfeWuHsZTMOM7j6j2+ooZiE4zQ1WFTcN+W0zdpX39EeufgzgoKJzWBckb3B/j5BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dc0baef332c94c8962e878ddba1d9e9092a542d23b5617dd5c8e9fa321da36d0","last_reissued_at":"2026-07-05T11:20:45.647442Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:45.647442Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Convergence Analysis of Natural Gradient Descent for Over-parameterized Physics-Informed Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bin Shan, Ting Du, Wang Kong, Xianliang Xu, Ye Li, Zhongyi Huang","submitted_at":"2024-08-01T14:06:34Z","abstract_excerpt":"In the context of over-parameterization, there is a line of work demonstrating that randomly initialized (stochastic) gradient descent (GD) converges to a globally optimal solution at a linear convergence rate for the quadratic loss function. However, the learning rate of GD for training two-layer neural networks exhibits poor dependence on the sample size and the Gram matrix, leading to a slow training process. In this paper, we show that for training two-layer $\\text{ReLU}^3$ Physics-Informed Neural Networks (PINNs), the learning rate can be improved from $\\mathcal{O}(\\lambda_0)$ to $\\mathca"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.00573","kind":"arxiv","version":4},"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/2408.00573/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":"2408.00573","created_at":"2026-07-05T11:20:45.647500+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.00573v4","created_at":"2026-07-05T11:20:45.647500+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.00573","created_at":"2026-07-05T11:20:45.647500+00:00"},{"alias_kind":"pith_short_12","alias_value":"3QF254ZSZFGI","created_at":"2026-07-05T11:20:45.647500+00:00"},{"alias_kind":"pith_short_16","alias_value":"3QF254ZSZFGISYXI","created_at":"2026-07-05T11:20:45.647500+00:00"},{"alias_kind":"pith_short_8","alias_value":"3QF254ZS","created_at":"2026-07-05T11:20:45.647500+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06369","citing_title":"Feature Learning for the High Dimensional Stationary Sch\\\"odinger Equation with Deep Ritz Method","ref_index":59,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC","json":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC.json","graph_json":"https://pith.science/api/pith-number/3QF254ZSZFGISYXIPDO3UHM6SC/graph.json","events_json":"https://pith.science/api/pith-number/3QF254ZSZFGISYXIPDO3UHM6SC/events.json","paper":"https://pith.science/paper/3QF254ZS"},"agent_actions":{"view_html":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC","download_json":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC.json","view_paper":"https://pith.science/paper/3QF254ZS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.00573&json=true","fetch_graph":"https://pith.science/api/pith-number/3QF254ZSZFGISYXIPDO3UHM6SC/graph.json","fetch_events":"https://pith.science/api/pith-number/3QF254ZSZFGISYXIPDO3UHM6SC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC/action/storage_attestation","attest_author":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC/action/author_attestation","sign_citation":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC/action/citation_signature","submit_replication":"https://pith.science/pith/3QF254ZSZFGISYXIPDO3UHM6SC/action/replication_record"}},"created_at":"2026-07-05T11:20:45.647500+00:00","updated_at":"2026-07-05T11:20:45.647500+00:00"}