{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:2U6BIEY37Q5JI7M5SYWVZ3DJXY","short_pith_number":"pith:2U6BIEY3","schema_version":"1.0","canonical_sha256":"d53c14131bfc3a947d9d962d5cec69be2398beada66612ef5f6e8fdef9ac7083","source":{"kind":"arxiv","id":"2006.10732","version":4},"attestation_state":"computed","paper":{"title":"When Does Preconditioning Help or Hurt Generalization?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Atsushi Nitanda, Denny Wu, Jimmy Ba, Ji Xu, Roger Grosse, Shun-ichi Amari, Taiji Suzuki, Xuechen Li","submitted_at":"2020-06-18T17:57:26Z","abstract_excerpt":"While second order optimizers such as natural gradient descent (NGD) often speed up optimization, their effect on generalization has been called into question. This work presents a more nuanced view on how the \\textit{implicit bias} of first- and second-order methods affects the comparison of generalization properties. We provide an exact asymptotic bias-variance decomposition of the generalization error of overparameterized ridgeless regression under a general class of preconditioner $\\boldsymbol{P}$, and consider the inverse population Fisher information matrix (used in NGD) as a particular "},"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":"2006.10732","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2020-06-18T17:57:26Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"84922202a3fc060826141175f9e90041a2e7caa4ec694d8fda44783170f71f4c","abstract_canon_sha256":"aeb06c9a26c03c13279dc0980ad6833218a2187262448a9fb5759fc3c6bfbc15"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:58:10.318187Z","signature_b64":"fIONXKh8uB8lkSX3Wa9bzgX3js8e+0e/kjHdbW3hQFJaee2ftAomdAGnktyghh0GLbblcnDBXc6Q0rs2ojOSDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d53c14131bfc3a947d9d962d5cec69be2398beada66612ef5f6e8fdef9ac7083","last_reissued_at":"2026-07-05T01:58:10.317646Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:58:10.317646Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Does Preconditioning Help or Hurt Generalization?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Atsushi Nitanda, Denny Wu, Jimmy Ba, Ji Xu, Roger Grosse, Shun-ichi Amari, Taiji Suzuki, Xuechen Li","submitted_at":"2020-06-18T17:57:26Z","abstract_excerpt":"While second order optimizers such as natural gradient descent (NGD) often speed up optimization, their effect on generalization has been called into question. This work presents a more nuanced view on how the \\textit{implicit bias} of first- and second-order methods affects the comparison of generalization properties. We provide an exact asymptotic bias-variance decomposition of the generalization error of overparameterized ridgeless regression under a general class of preconditioner $\\boldsymbol{P}$, and consider the inverse population Fisher information matrix (used in NGD) as a particular "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2006.10732","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/2006.10732/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":"2006.10732","created_at":"2026-07-05T01:58:10.317710+00:00"},{"alias_kind":"arxiv_version","alias_value":"2006.10732v4","created_at":"2026-07-05T01:58:10.317710+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2006.10732","created_at":"2026-07-05T01:58:10.317710+00:00"},{"alias_kind":"pith_short_12","alias_value":"2U6BIEY37Q5J","created_at":"2026-07-05T01:58:10.317710+00:00"},{"alias_kind":"pith_short_16","alias_value":"2U6BIEY37Q5JI7M5","created_at":"2026-07-05T01:58:10.317710+00:00"},{"alias_kind":"pith_short_8","alias_value":"2U6BIEY3","created_at":"2026-07-05T01:58:10.317710+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2502.05075","citing_title":"Discrepancies are Virtue: Weak-to-Strong Generalization through Lens of Intrinsic Dimension","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18591","citing_title":"Randomized Advantage Transformation (RAT): Computing Natural Policy Gradients via Direct Backpropagation","ref_index":106,"is_internal_anchor":false},{"citing_arxiv_id":"2601.03162","citing_title":"On the Convergence Behavior of Preconditioned Gradient Descent Toward the Rich Learning Regime","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09471","citing_title":"The Statistical Cost of Adaptation in Multi-Source Transfer Learning","ref_index":160,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY","json":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY.json","graph_json":"https://pith.science/api/pith-number/2U6BIEY37Q5JI7M5SYWVZ3DJXY/graph.json","events_json":"https://pith.science/api/pith-number/2U6BIEY37Q5JI7M5SYWVZ3DJXY/events.json","paper":"https://pith.science/paper/2U6BIEY3"},"agent_actions":{"view_html":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY","download_json":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY.json","view_paper":"https://pith.science/paper/2U6BIEY3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2006.10732&json=true","fetch_graph":"https://pith.science/api/pith-number/2U6BIEY37Q5JI7M5SYWVZ3DJXY/graph.json","fetch_events":"https://pith.science/api/pith-number/2U6BIEY37Q5JI7M5SYWVZ3DJXY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY/action/storage_attestation","attest_author":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY/action/author_attestation","sign_citation":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY/action/citation_signature","submit_replication":"https://pith.science/pith/2U6BIEY37Q5JI7M5SYWVZ3DJXY/action/replication_record"}},"created_at":"2026-07-05T01:58:10.317710+00:00","updated_at":"2026-07-05T01:58:10.317710+00:00"}