{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:PT6BJ2JTYTYCZPCTOC62IX5CPK","short_pith_number":"pith:PT6BJ2JT","schema_version":"1.0","canonical_sha256":"7cfc14e933c4f02cbc5370bda45fa27ab28590ca72e7250643ccc104330add83","source":{"kind":"arxiv","id":"1806.01796","version":3},"attestation_state":"computed","paper":{"title":"Stochastic Gradient Descent on Separable Data: Exact Convergence with a Fixed Learning Rate","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Daniel Soudry, Mor Shpigel Nacson, Nathan Srebro","submitted_at":"2018-06-05T16:37:19Z","abstract_excerpt":"Stochastic Gradient Descent (SGD) is a central tool in machine learning. We prove that SGD converges to zero loss, even with a fixed (non-vanishing) learning rate - in the special case of homogeneous linear classifiers with smooth monotone loss functions, optimized on linearly separable data. Previous works assumed either a vanishing learning rate, iterate averaging, or loss assumptions that do not hold for monotone loss functions used for classification, such as the logistic loss. We prove our result on a fixed dataset, both for sampling with or without replacement. Furthermore, for logistic "},"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":"1806.01796","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2018-06-05T16:37:19Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"4bcebdcac6808e3a334fe3e532dd60d0037af701089982942c44b1877a69a620","abstract_canon_sha256":"c4fcb7e09938c25d3bec397e34988c5288d171cc53c5d51b4290f76d17312209"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:15:05.780196Z","signature_b64":"h1hYnrJM6VEoTlOPWIxfqvFarzT0l154F2D7vKeZj3pjCHi0pvYWPp2NQ61mP9Y18vK/hG6XB+fxuTrvk9doDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7cfc14e933c4f02cbc5370bda45fa27ab28590ca72e7250643ccc104330add83","last_reissued_at":"2026-07-05T04:15:05.779741Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:15:05.779741Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stochastic Gradient Descent on Separable Data: Exact Convergence with a Fixed Learning Rate","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Daniel Soudry, Mor Shpigel Nacson, Nathan Srebro","submitted_at":"2018-06-05T16:37:19Z","abstract_excerpt":"Stochastic Gradient Descent (SGD) is a central tool in machine learning. We prove that SGD converges to zero loss, even with a fixed (non-vanishing) learning rate - in the special case of homogeneous linear classifiers with smooth monotone loss functions, optimized on linearly separable data. Previous works assumed either a vanishing learning rate, iterate averaging, or loss assumptions that do not hold for monotone loss functions used for classification, such as the logistic loss. We prove our result on a fixed dataset, both for sampling with or without replacement. Furthermore, for logistic "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1806.01796","kind":"arxiv","version":3},"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/1806.01796/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":"1806.01796","created_at":"2026-07-05T04:15:05.779798+00:00"},{"alias_kind":"arxiv_version","alias_value":"1806.01796v3","created_at":"2026-07-05T04:15:05.779798+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1806.01796","created_at":"2026-07-05T04:15:05.779798+00:00"},{"alias_kind":"pith_short_12","alias_value":"PT6BJ2JTYTYC","created_at":"2026-07-05T04:15:05.779798+00:00"},{"alias_kind":"pith_short_16","alias_value":"PT6BJ2JTYTYCZPCT","created_at":"2026-07-05T04:15:05.779798+00:00"},{"alias_kind":"pith_short_8","alias_value":"PT6BJ2JT","created_at":"2026-07-05T04:15:05.779798+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.23489","citing_title":"SGD as Free Energy Minimization: A Thermodynamic View on Neural Network Training","ref_index":45,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK","json":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK.json","graph_json":"https://pith.science/api/pith-number/PT6BJ2JTYTYCZPCTOC62IX5CPK/graph.json","events_json":"https://pith.science/api/pith-number/PT6BJ2JTYTYCZPCTOC62IX5CPK/events.json","paper":"https://pith.science/paper/PT6BJ2JT"},"agent_actions":{"view_html":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK","download_json":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK.json","view_paper":"https://pith.science/paper/PT6BJ2JT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1806.01796&json=true","fetch_graph":"https://pith.science/api/pith-number/PT6BJ2JTYTYCZPCTOC62IX5CPK/graph.json","fetch_events":"https://pith.science/api/pith-number/PT6BJ2JTYTYCZPCTOC62IX5CPK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK/action/storage_attestation","attest_author":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK/action/author_attestation","sign_citation":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK/action/citation_signature","submit_replication":"https://pith.science/pith/PT6BJ2JTYTYCZPCTOC62IX5CPK/action/replication_record"}},"created_at":"2026-07-05T04:15:05.779798+00:00","updated_at":"2026-07-05T04:15:05.779798+00:00"}