{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:DETQEFSCV5K4FOKGAC5NBSA4C6","short_pith_number":"pith:DETQEFSC","schema_version":"1.0","canonical_sha256":"1927021642af55c2b94600bad0c81c17bf1541fcead5e261b327f1c2c271dc43","source":{"kind":"arxiv","id":"2511.19716","version":2},"attestation_state":"computed","paper":{"title":"Design Criteria for SGD Preconditioners: Local Conditioning, Noise Floors, and Basin Stability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","cs.NA"],"primary_cat":"math.NA","authors_text":"Alexandra Pichette-Emmons, Mitchell Scott, Qiang Ye, Tianshi Xu, Yousef Saad, Yuanzhe Xi, Ziyuan Tang","submitted_at":"2025-11-24T21:24:40Z","abstract_excerpt":"Stochastic Gradient Descent (SGD) often slows in the late stage of training due to anisotropic curvature and gradient noise. We analyze preconditioned SGD in the geometry induced by a symmetric positive definite matrix $\\mathbf{M}$, deriving bounds in which both the convergence rate and the stochastic noise floor are governed by $\\mathbf{M}$-dependent quantities: the rate through an effective condition number in the $\\mathbf{M}$-metric, and the floor through the product of that condition number and the preconditioned noise level. For nonconvex objectives, we establish a preconditioner-dependen"},"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":"2511.19716","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"math.NA","submitted_at":"2025-11-24T21:24:40Z","cross_cats_sorted":["cs.LG","cs.NA"],"title_canon_sha256":"fcfd1a9b0ca60bfdcee01a6ebe3bffeeb492059d32f16b560a809346fc2cfeb1","abstract_canon_sha256":"175fc53f5f0b9e5754cb7b0b3b08e96ce6513fde2cb3c53ed2d010795e58e050"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-12T01:08:18.536234Z","signature_b64":"d8KItkfRINKV+JNvWvrY65A8HT7Jati2QBjeoeqoRo6WDyUdNCczrrH9jZWztYKOl0VJRRsg3VaO1oXoO43gDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1927021642af55c2b94600bad0c81c17bf1541fcead5e261b327f1c2c271dc43","last_reissued_at":"2026-06-12T01:08:18.535204Z","signature_status":"signed_v1","first_computed_at":"2026-06-12T01:08:18.535204Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Design Criteria for SGD Preconditioners: Local Conditioning, Noise Floors, and Basin Stability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","cs.NA"],"primary_cat":"math.NA","authors_text":"Alexandra Pichette-Emmons, Mitchell Scott, Qiang Ye, Tianshi Xu, Yousef Saad, Yuanzhe Xi, Ziyuan Tang","submitted_at":"2025-11-24T21:24:40Z","abstract_excerpt":"Stochastic Gradient Descent (SGD) often slows in the late stage of training due to anisotropic curvature and gradient noise. We analyze preconditioned SGD in the geometry induced by a symmetric positive definite matrix $\\mathbf{M}$, deriving bounds in which both the convergence rate and the stochastic noise floor are governed by $\\mathbf{M}$-dependent quantities: the rate through an effective condition number in the $\\mathbf{M}$-metric, and the floor through the product of that condition number and the preconditioned noise level. For nonconvex objectives, we establish a preconditioner-dependen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2511.19716","kind":"arxiv","version":2},"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/2511.19716/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":"2511.19716","created_at":"2026-06-12T01:08:18.535377+00:00"},{"alias_kind":"arxiv_version","alias_value":"2511.19716v2","created_at":"2026-06-12T01:08:18.535377+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.19716","created_at":"2026-06-12T01:08:18.535377+00:00"},{"alias_kind":"pith_short_12","alias_value":"DETQEFSCV5K4","created_at":"2026-06-12T01:08:18.535377+00:00"},{"alias_kind":"pith_short_16","alias_value":"DETQEFSCV5K4FOKG","created_at":"2026-06-12T01:08:18.535377+00:00"},{"alias_kind":"pith_short_8","alias_value":"DETQEFSC","created_at":"2026-06-12T01:08:18.535377+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.23498","citing_title":"When Does Dynamic Preconditioning Preserve the Polyak-Ruppert CLT? A Stabilization Threshold","ref_index":35,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6","json":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6.json","graph_json":"https://pith.science/api/pith-number/DETQEFSCV5K4FOKGAC5NBSA4C6/graph.json","events_json":"https://pith.science/api/pith-number/DETQEFSCV5K4FOKGAC5NBSA4C6/events.json","paper":"https://pith.science/paper/DETQEFSC"},"agent_actions":{"view_html":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6","download_json":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6.json","view_paper":"https://pith.science/paper/DETQEFSC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2511.19716&json=true","fetch_graph":"https://pith.science/api/pith-number/DETQEFSCV5K4FOKGAC5NBSA4C6/graph.json","fetch_events":"https://pith.science/api/pith-number/DETQEFSCV5K4FOKGAC5NBSA4C6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6/action/storage_attestation","attest_author":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6/action/author_attestation","sign_citation":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6/action/citation_signature","submit_replication":"https://pith.science/pith/DETQEFSCV5K4FOKGAC5NBSA4C6/action/replication_record"}},"created_at":"2026-06-12T01:08:18.535377+00:00","updated_at":"2026-06-12T01:08:18.535377+00:00"}