{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:JZS2MP3CGA5MN54W3GBYT5MMZH","short_pith_number":"pith:JZS2MP3C","schema_version":"1.0","canonical_sha256":"4e65a63f62303ac6f796d98389f58cc9f2bf780cae6bf50d7d260ab6440e76db","source":{"kind":"arxiv","id":"2310.07831","version":2},"attestation_state":"computed","paper":{"title":"Optimal Linear Decay Learning Rate Schedules and Further Refinements","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aaron Defazio, Ashok Cutkosky, Harsh Mehta, Konstantin Mishchenko","submitted_at":"2023-10-11T19:16:35Z","abstract_excerpt":"Learning rate schedules used in practice bear little resemblance to those recommended by theory. We close much of this theory/practice gap, and as a consequence are able to derive new problem-adaptive learning rate schedules. Our main technical contribution is a refined analysis of learning rate schedules for a wide class of optimization algorithms (including SGD). When considering only worst-case analysis, our theory predicts that the optimal choice is the linear decay schedule where the step-size is set proportional to 1 - t/T, where t is the current iteration and T is the total number of st"},"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":"2310.07831","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-11T19:16:35Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"498a187027c6a815409477cb36d00ec5349faa9e68d5816cebf1f550d20592aa","abstract_canon_sha256":"106fa98131f18ef3b2ef73b1c7c9ca453ed1face29f9bee8d4a5baa19285d8c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:28:09.560461Z","signature_b64":"R1oWWqs5nL/C1Taxck625I+9O4QKK9uKxMtBjRgC2ryRD4PfWk75Exd3+LJwrHJcgAlVNPsgz0vqXU0M1le7CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4e65a63f62303ac6f796d98389f58cc9f2bf780cae6bf50d7d260ab6440e76db","last_reissued_at":"2026-07-05T09:28:09.559987Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:28:09.559987Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimal Linear Decay Learning Rate Schedules and Further Refinements","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aaron Defazio, Ashok Cutkosky, Harsh Mehta, Konstantin Mishchenko","submitted_at":"2023-10-11T19:16:35Z","abstract_excerpt":"Learning rate schedules used in practice bear little resemblance to those recommended by theory. We close much of this theory/practice gap, and as a consequence are able to derive new problem-adaptive learning rate schedules. Our main technical contribution is a refined analysis of learning rate schedules for a wide class of optimization algorithms (including SGD). When considering only worst-case analysis, our theory predicts that the optimal choice is the linear decay schedule where the step-size is set proportional to 1 - t/T, where t is the current iteration and T is the total number of st"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.07831","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/2310.07831/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":"2310.07831","created_at":"2026-07-05T09:28:09.560042+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.07831v2","created_at":"2026-07-05T09:28:09.560042+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.07831","created_at":"2026-07-05T09:28:09.560042+00:00"},{"alias_kind":"pith_short_12","alias_value":"JZS2MP3CGA5M","created_at":"2026-07-05T09:28:09.560042+00:00"},{"alias_kind":"pith_short_16","alias_value":"JZS2MP3CGA5MN54W","created_at":"2026-07-05T09:28:09.560042+00:00"},{"alias_kind":"pith_short_8","alias_value":"JZS2MP3C","created_at":"2026-07-05T09:28:09.560042+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.30559","citing_title":"Convergence of Continual Learning in Homogeneous Deep Networks","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2511.07767","citing_title":"Taking the Road Less Scheduled with Adaptive Polyak Steps","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2602.04774","citing_title":"Theory of Optimal Learning Rate Schedules and Scaling Laws for a Random Feature Model","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10728","citing_title":"Last Iterate Convergence of AdaGrad-Norm for Convex Non-Smooth Optimization","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06741","citing_title":"A Closed-Form Upper Bound for Admissible Learning-Rate Steps in Belief-Space Dynamics","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH","json":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH.json","graph_json":"https://pith.science/api/pith-number/JZS2MP3CGA5MN54W3GBYT5MMZH/graph.json","events_json":"https://pith.science/api/pith-number/JZS2MP3CGA5MN54W3GBYT5MMZH/events.json","paper":"https://pith.science/paper/JZS2MP3C"},"agent_actions":{"view_html":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH","download_json":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH.json","view_paper":"https://pith.science/paper/JZS2MP3C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.07831&json=true","fetch_graph":"https://pith.science/api/pith-number/JZS2MP3CGA5MN54W3GBYT5MMZH/graph.json","fetch_events":"https://pith.science/api/pith-number/JZS2MP3CGA5MN54W3GBYT5MMZH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH/action/storage_attestation","attest_author":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH/action/author_attestation","sign_citation":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH/action/citation_signature","submit_replication":"https://pith.science/pith/JZS2MP3CGA5MN54W3GBYT5MMZH/action/replication_record"}},"created_at":"2026-07-05T09:28:09.560042+00:00","updated_at":"2026-07-05T09:28:09.560042+00:00"}