{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:4K6GSOJWB7QU2RFIL3CD3BQJ62","short_pith_number":"pith:4K6GSOJW","schema_version":"1.0","canonical_sha256":"e2bc6939360fe14d44a85ec43d8609f6bf468456709f3f2ee52fa1625102ed2a","source":{"kind":"arxiv","id":"2106.06530","version":2},"attestation_state":"computed","paper":{"title":"Label Noise SGD Provably Prefers Flat Global Minimizers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IT","math.IT","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Alex Damian, Jason D. Lee, Tengyu Ma","submitted_at":"2021-06-11T17:59:07Z","abstract_excerpt":"In overparametrized models, the noise in stochastic gradient descent (SGD) implicitly regularizes the optimization trajectory and determines which local minimum SGD converges to. Motivated by empirical studies that demonstrate that training with noisy labels improves generalization, we study the implicit regularization effect of SGD with label noise. We show that SGD with label noise converges to a stationary point of a regularized loss $L(\\theta) +\\lambda R(\\theta)$, where $L(\\theta)$ is the training loss, $\\lambda$ is an effective regularization parameter depending on the step size, strength"},"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":"2106.06530","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-11T17:59:07Z","cross_cats_sorted":["cs.IT","math.IT","math.OC","stat.ML"],"title_canon_sha256":"864e0ef86f4653ff5b2a4abf1be7414173777e97bd43d65862af74616d42ff1c","abstract_canon_sha256":"ffdf6e2db75a3f3df013f49c45c3a9c5b5ebee471d4d6175a95445fe0acc449f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:37:43.391499Z","signature_b64":"VwcviEJLU0pz+MoyH+70OORhNxQT0ROQqN6ZS6xbmtU7VsCDV7WXdlwJuNwQqgFX+GTIIObwYpYpufZJjU1NDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e2bc6939360fe14d44a85ec43d8609f6bf468456709f3f2ee52fa1625102ed2a","last_reissued_at":"2026-07-05T03:37:43.391067Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:37:43.391067Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Label Noise SGD Provably Prefers Flat Global Minimizers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IT","math.IT","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Alex Damian, Jason D. Lee, Tengyu Ma","submitted_at":"2021-06-11T17:59:07Z","abstract_excerpt":"In overparametrized models, the noise in stochastic gradient descent (SGD) implicitly regularizes the optimization trajectory and determines which local minimum SGD converges to. Motivated by empirical studies that demonstrate that training with noisy labels improves generalization, we study the implicit regularization effect of SGD with label noise. We show that SGD with label noise converges to a stationary point of a regularized loss $L(\\theta) +\\lambda R(\\theta)$, where $L(\\theta)$ is the training loss, $\\lambda$ is an effective regularization parameter depending on the step size, strength"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.06530","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/2106.06530/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":"2106.06530","created_at":"2026-07-05T03:37:43.391131+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.06530v2","created_at":"2026-07-05T03:37:43.391131+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.06530","created_at":"2026-07-05T03:37:43.391131+00:00"},{"alias_kind":"pith_short_12","alias_value":"4K6GSOJWB7QU","created_at":"2026-07-05T03:37:43.391131+00:00"},{"alias_kind":"pith_short_16","alias_value":"4K6GSOJWB7QU2RFI","created_at":"2026-07-05T03:37:43.391131+00:00"},{"alias_kind":"pith_short_8","alias_value":"4K6GSOJW","created_at":"2026-07-05T03:37:43.391131+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62","json":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62.json","graph_json":"https://pith.science/api/pith-number/4K6GSOJWB7QU2RFIL3CD3BQJ62/graph.json","events_json":"https://pith.science/api/pith-number/4K6GSOJWB7QU2RFIL3CD3BQJ62/events.json","paper":"https://pith.science/paper/4K6GSOJW"},"agent_actions":{"view_html":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62","download_json":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62.json","view_paper":"https://pith.science/paper/4K6GSOJW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.06530&json=true","fetch_graph":"https://pith.science/api/pith-number/4K6GSOJWB7QU2RFIL3CD3BQJ62/graph.json","fetch_events":"https://pith.science/api/pith-number/4K6GSOJWB7QU2RFIL3CD3BQJ62/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62/action/storage_attestation","attest_author":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62/action/author_attestation","sign_citation":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62/action/citation_signature","submit_replication":"https://pith.science/pith/4K6GSOJWB7QU2RFIL3CD3BQJ62/action/replication_record"}},"created_at":"2026-07-05T03:37:43.391131+00:00","updated_at":"2026-07-05T03:37:43.391131+00:00"}