{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KNA3GHANMLOU4KEDYOPGM5WCQ7","short_pith_number":"pith:KNA3GHAN","schema_version":"1.0","canonical_sha256":"5341b31c0d62dd4e2883c39e6676c287fd8bd6ed9e208a4f0ea4e3cc0b725b35","source":{"kind":"arxiv","id":"2406.09405","version":2},"attestation_state":"computed","paper":{"title":"Why Warmup the Learning Rate? Underlying Mechanisms and Improvements","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","stat.ML"],"primary_cat":"cs.LG","authors_text":"Dayal Singh Kalra, Maissam Barkeshli","submitted_at":"2024-06-13T17:59:35Z","abstract_excerpt":"It is common in deep learning to warm up the learning rate $\\eta$, often by a linear schedule between $\\eta_{\\text{init}} = 0$ and a predetermined target $\\eta_{\\text{trgt}}$. In this paper, we show through systematic experiments using SGD and Adam that the overwhelming benefit of warmup arises from allowing the network to tolerate larger $\\eta_{\\text{trgt}}$ {by forcing the network to more well-conditioned areas of the loss landscape}. The ability to handle larger $\\eta_{\\text{trgt}}$ makes hyperparameter tuning more robust while improving the final performance. We uncover different regimes o"},"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":"2406.09405","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-13T17:59:35Z","cross_cats_sorted":["cond-mat.dis-nn","stat.ML"],"title_canon_sha256":"3804dc8bf442ca94025972cb158abd329ebaca4f20aacbae218c94ca3c646cf2","abstract_canon_sha256":"95583cb9c0059913f8755d1175da50a716e147443a3a5c9999cbadca3456a452"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:29:49.562516Z","signature_b64":"h2sRsTnZvn3Tv4tk6Gk0g6x0mVlIZQP3/dME0e5qAoUq7n74/dI/A99JH+9GcUoDruSG/LBvYvULE+WmLPHhBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5341b31c0d62dd4e2883c39e6676c287fd8bd6ed9e208a4f0ea4e3cc0b725b35","last_reissued_at":"2026-07-05T09:29:49.562019Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:29:49.562019Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Why Warmup the Learning Rate? Underlying Mechanisms and Improvements","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","stat.ML"],"primary_cat":"cs.LG","authors_text":"Dayal Singh Kalra, Maissam Barkeshli","submitted_at":"2024-06-13T17:59:35Z","abstract_excerpt":"It is common in deep learning to warm up the learning rate $\\eta$, often by a linear schedule between $\\eta_{\\text{init}} = 0$ and a predetermined target $\\eta_{\\text{trgt}}$. In this paper, we show through systematic experiments using SGD and Adam that the overwhelming benefit of warmup arises from allowing the network to tolerate larger $\\eta_{\\text{trgt}}$ {by forcing the network to more well-conditioned areas of the loss landscape}. The ability to handle larger $\\eta_{\\text{trgt}}$ makes hyperparameter tuning more robust while improving the final performance. We uncover different regimes o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.09405","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/2406.09405/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":"2406.09405","created_at":"2026-07-05T09:29:49.562079+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.09405v2","created_at":"2026-07-05T09:29:49.562079+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.09405","created_at":"2026-07-05T09:29:49.562079+00:00"},{"alias_kind":"pith_short_12","alias_value":"KNA3GHANMLOU","created_at":"2026-07-05T09:29:49.562079+00:00"},{"alias_kind":"pith_short_16","alias_value":"KNA3GHANMLOU4KED","created_at":"2026-07-05T09:29:49.562079+00:00"},{"alias_kind":"pith_short_8","alias_value":"KNA3GHAN","created_at":"2026-07-05T09:29:49.562079+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.03998","citing_title":"Directional Curvature from Armijo Backtracking: A Low-Cost Sharpness Probe and a Calibration-Free Learning-Rate Safeguard for Adam","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7","json":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7.json","graph_json":"https://pith.science/api/pith-number/KNA3GHANMLOU4KEDYOPGM5WCQ7/graph.json","events_json":"https://pith.science/api/pith-number/KNA3GHANMLOU4KEDYOPGM5WCQ7/events.json","paper":"https://pith.science/paper/KNA3GHAN"},"agent_actions":{"view_html":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7","download_json":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7.json","view_paper":"https://pith.science/paper/KNA3GHAN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.09405&json=true","fetch_graph":"https://pith.science/api/pith-number/KNA3GHANMLOU4KEDYOPGM5WCQ7/graph.json","fetch_events":"https://pith.science/api/pith-number/KNA3GHANMLOU4KEDYOPGM5WCQ7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7/action/storage_attestation","attest_author":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7/action/author_attestation","sign_citation":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7/action/citation_signature","submit_replication":"https://pith.science/pith/KNA3GHANMLOU4KEDYOPGM5WCQ7/action/replication_record"}},"created_at":"2026-07-05T09:29:49.562079+00:00","updated_at":"2026-07-05T09:29:49.562079+00:00"}