{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TAVFBZ5NM6VK2EEMPWCGZLELTC","short_pith_number":"pith:TAVFBZ5N","schema_version":"1.0","canonical_sha256":"982a50e7ad67aaad108c7d846cac8b98b8f25848b8cffac30a55504780e20fd4","source":{"kind":"arxiv","id":"2504.04105","version":2},"attestation_state":"computed","paper":{"title":"Minimax Optimal Convergence of Gradient Descent in Logistic Regression via Large and Adaptive Stepsizes","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Jingfeng Wu, Licong Lin, Peter L. Bartlett, Ruiqi Zhang","submitted_at":"2025-04-05T08:34:20Z","abstract_excerpt":"We study $\\textit{gradient descent}$ (GD) for logistic regression on linearly separable data with stepsizes that adapt to the current risk, scaled by a constant hyperparameter $\\eta$. We show that after at most $1/\\gamma^2$ burn-in steps, GD achieves a risk upper bounded by $\\exp(-\\Theta(\\eta))$, where $\\gamma$ is the margin of the dataset. As $\\eta$ can be arbitrarily large, GD attains an arbitrarily small risk $\\textit{immediately after the burn-in steps}$, though the risk evolution may be $\\textit{non-monotonic}$.\n  We further construct hard datasets with margin $\\gamma$, where any batch (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":"2504.04105","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2025-04-05T08:34:20Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"28f3e56d5b3e1a6d35992573bf616ceab2862ee7d2ee1fa6ddd03c1447face81","abstract_canon_sha256":"36b15aba65800679ee6b8fcaad36c598cb34c6f8546e68ba48d7e835703fe304"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:50:48.349907Z","signature_b64":"IplD2+2gDN41jqmqHX2xrxoacFRj2V4nt8WH7lYugvbjvS1cpYk5JTAm5asdFyaUa4JB0CufeSTtBsxDJhVhAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"982a50e7ad67aaad108c7d846cac8b98b8f25848b8cffac30a55504780e20fd4","last_reissued_at":"2026-07-05T10:50:48.349378Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:50:48.349378Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Minimax Optimal Convergence of Gradient Descent in Logistic Regression via Large and Adaptive Stepsizes","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Jingfeng Wu, Licong Lin, Peter L. Bartlett, Ruiqi Zhang","submitted_at":"2025-04-05T08:34:20Z","abstract_excerpt":"We study $\\textit{gradient descent}$ (GD) for logistic regression on linearly separable data with stepsizes that adapt to the current risk, scaled by a constant hyperparameter $\\eta$. We show that after at most $1/\\gamma^2$ burn-in steps, GD achieves a risk upper bounded by $\\exp(-\\Theta(\\eta))$, where $\\gamma$ is the margin of the dataset. As $\\eta$ can be arbitrarily large, GD attains an arbitrarily small risk $\\textit{immediately after the burn-in steps}$, though the risk evolution may be $\\textit{non-monotonic}$.\n  We further construct hard datasets with margin $\\gamma$, where any batch (o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.04105","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/2504.04105/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":"2504.04105","created_at":"2026-07-05T10:50:48.349440+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.04105v2","created_at":"2026-07-05T10:50:48.349440+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.04105","created_at":"2026-07-05T10:50:48.349440+00:00"},{"alias_kind":"pith_short_12","alias_value":"TAVFBZ5NM6VK","created_at":"2026-07-05T10:50:48.349440+00:00"},{"alias_kind":"pith_short_16","alias_value":"TAVFBZ5NM6VK2EEM","created_at":"2026-07-05T10:50:48.349440+00:00"},{"alias_kind":"pith_short_8","alias_value":"TAVFBZ5N","created_at":"2026-07-05T10:50:48.349440+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21683","citing_title":"Finite-Sample Performance of Gradient Descent in Logistic Regression with Gaussian Design","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21292","citing_title":"Large-Step Training Dynamics of a Two-Factor Linear Transformer Model","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14200","citing_title":"How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02920","citing_title":"Efficient Logistic Regression with Mixture of Sigmoids","ref_index":47,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC","json":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC.json","graph_json":"https://pith.science/api/pith-number/TAVFBZ5NM6VK2EEMPWCGZLELTC/graph.json","events_json":"https://pith.science/api/pith-number/TAVFBZ5NM6VK2EEMPWCGZLELTC/events.json","paper":"https://pith.science/paper/TAVFBZ5N"},"agent_actions":{"view_html":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC","download_json":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC.json","view_paper":"https://pith.science/paper/TAVFBZ5N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.04105&json=true","fetch_graph":"https://pith.science/api/pith-number/TAVFBZ5NM6VK2EEMPWCGZLELTC/graph.json","fetch_events":"https://pith.science/api/pith-number/TAVFBZ5NM6VK2EEMPWCGZLELTC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC/action/storage_attestation","attest_author":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC/action/author_attestation","sign_citation":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC/action/citation_signature","submit_replication":"https://pith.science/pith/TAVFBZ5NM6VK2EEMPWCGZLELTC/action/replication_record"}},"created_at":"2026-07-05T10:50:48.349440+00:00","updated_at":"2026-07-05T10:50:48.349440+00:00"}