{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:3URAVUCD7HQOXW7XIKSCD3VW4O","short_pith_number":"pith:3URAVUCD","schema_version":"1.0","canonical_sha256":"dd220ad043f9e0ebdbf742a421eeb6e3b1a7a9796cdad8fbc0783055f71f2441","source":{"kind":"arxiv","id":"2403.04081","version":2},"attestation_state":"computed","paper":{"title":"Directional Smoothness and Gradient Methods: Convergence and Adaptivity","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Aaron Defazio, Aaron Mishkin, Ahmed Khaled, Robert M. Gower, Yuanhao Wang","submitted_at":"2024-03-06T22:24:05Z","abstract_excerpt":"We develop new sub-optimality bounds for gradient descent (GD) that depend on the conditioning of the objective along the path of optimization rather than on global, worst-case constants. Key to our proofs is directional smoothness, a measure of gradient variation that we use to develop upper-bounds on the objective. Minimizing these upper-bounds requires solving implicit equations to obtain a sequence of strongly adapted step-sizes; we show that these equations are straightforward to solve for convex quadratics and lead to new guarantees for two classical step-sizes. For general functions, we"},"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":"2403.04081","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-03-06T22:24:05Z","cross_cats_sorted":["math.OC"],"title_canon_sha256":"5a1d4ca13b5eb88060b28be96c883559019b153d64820870239a2e76f219ee95","abstract_canon_sha256":"acac133b4d65d3da883e97e393ddb1ec7a3bc9b895b053d6f400dd1c7db593e3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:00:35.132909Z","signature_b64":"5ylkjmM/X7BKuoJZTQl+XVyA316b45fu6Asiy+OMU0qUraW1T8WH/mx5JpQJ1JXywnyxZiV6mrEWek16RQgoBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dd220ad043f9e0ebdbf742a421eeb6e3b1a7a9796cdad8fbc0783055f71f2441","last_reissued_at":"2026-07-05T10:00:35.132514Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:00:35.132514Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Directional Smoothness and Gradient Methods: Convergence and Adaptivity","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Aaron Defazio, Aaron Mishkin, Ahmed Khaled, Robert M. Gower, Yuanhao Wang","submitted_at":"2024-03-06T22:24:05Z","abstract_excerpt":"We develop new sub-optimality bounds for gradient descent (GD) that depend on the conditioning of the objective along the path of optimization rather than on global, worst-case constants. Key to our proofs is directional smoothness, a measure of gradient variation that we use to develop upper-bounds on the objective. Minimizing these upper-bounds requires solving implicit equations to obtain a sequence of strongly adapted step-sizes; we show that these equations are straightforward to solve for convex quadratics and lead to new guarantees for two classical step-sizes. For general functions, we"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.04081","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/2403.04081/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":"2403.04081","created_at":"2026-07-05T10:00:35.132571+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.04081v2","created_at":"2026-07-05T10:00:35.132571+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.04081","created_at":"2026-07-05T10:00:35.132571+00:00"},{"alias_kind":"pith_short_12","alias_value":"3URAVUCD7HQO","created_at":"2026-07-05T10:00:35.132571+00:00"},{"alias_kind":"pith_short_16","alias_value":"3URAVUCD7HQOXW7X","created_at":"2026-07-05T10:00:35.132571+00:00"},{"alias_kind":"pith_short_8","alias_value":"3URAVUCD","created_at":"2026-07-05T10:00:35.132571+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.03701","citing_title":"First-ish Order Methods: Hessian-aware Scalings of Gradient Descent","ref_index":50,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O","json":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O.json","graph_json":"https://pith.science/api/pith-number/3URAVUCD7HQOXW7XIKSCD3VW4O/graph.json","events_json":"https://pith.science/api/pith-number/3URAVUCD7HQOXW7XIKSCD3VW4O/events.json","paper":"https://pith.science/paper/3URAVUCD"},"agent_actions":{"view_html":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O","download_json":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O.json","view_paper":"https://pith.science/paper/3URAVUCD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.04081&json=true","fetch_graph":"https://pith.science/api/pith-number/3URAVUCD7HQOXW7XIKSCD3VW4O/graph.json","fetch_events":"https://pith.science/api/pith-number/3URAVUCD7HQOXW7XIKSCD3VW4O/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O/action/storage_attestation","attest_author":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O/action/author_attestation","sign_citation":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O/action/citation_signature","submit_replication":"https://pith.science/pith/3URAVUCD7HQOXW7XIKSCD3VW4O/action/replication_record"}},"created_at":"2026-07-05T10:00:35.132571+00:00","updated_at":"2026-07-05T10:00:35.132571+00:00"}