{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TYQTYRV6WK73YLIUZJ4PFODT6K","short_pith_number":"pith:TYQTYRV6","schema_version":"1.0","canonical_sha256":"9e213c46beb2bfbc2d14ca78f2b873f28f1859bd127c18e82a3fe03540d25d84","source":{"kind":"arxiv","id":"2505.08306","version":1},"attestation_state":"computed","paper":{"title":"Rapid Overfitting of Multi-Pass Stochastic Gradient Descent in Stochastic Convex Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Roi Livni, Shira Vansover-Hager, Tomer Koren","submitted_at":"2025-05-13T07:32:48Z","abstract_excerpt":"We study the out-of-sample performance of multi-pass stochastic gradient descent (SGD) in the fundamental stochastic convex optimization (SCO) model. While one-pass SGD is known to achieve an optimal $\\Theta(1/\\sqrt{n})$ excess population loss given a sample of size $n$, much less is understood about the multi-pass version of the algorithm which is widely used in practice. Somewhat surprisingly, we show that in the general non-smooth case of SCO, just a few epochs of SGD can already hurt its out-of-sample performance significantly and lead to overfitting. In particular, using a step size $\\eta"},"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":"2505.08306","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-05-13T07:32:48Z","cross_cats_sorted":["math.OC","stat.ML"],"title_canon_sha256":"08194322843bfc0463b026cc5997447975f1e80545e984d5b8c93601c34b7b45","abstract_canon_sha256":"7b29513b9d97e48a73c74ef4a62409cab3fc8a524774ca22106c1dcae8237791"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:03:24.783481Z","signature_b64":"BWNnArgiXyAjjdb5tdQ2+P+MDIG13ImU+iNoPq27vTRg140/Mo2bKCl0kvbjd8rRy/J7TeXymQVtykrJM/g4Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9e213c46beb2bfbc2d14ca78f2b873f28f1859bd127c18e82a3fe03540d25d84","last_reissued_at":"2026-07-05T11:03:24.783030Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:03:24.783030Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rapid Overfitting of Multi-Pass Stochastic Gradient Descent in Stochastic Convex Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Roi Livni, Shira Vansover-Hager, Tomer Koren","submitted_at":"2025-05-13T07:32:48Z","abstract_excerpt":"We study the out-of-sample performance of multi-pass stochastic gradient descent (SGD) in the fundamental stochastic convex optimization (SCO) model. While one-pass SGD is known to achieve an optimal $\\Theta(1/\\sqrt{n})$ excess population loss given a sample of size $n$, much less is understood about the multi-pass version of the algorithm which is widely used in practice. Somewhat surprisingly, we show that in the general non-smooth case of SCO, just a few epochs of SGD can already hurt its out-of-sample performance significantly and lead to overfitting. In particular, using a step size $\\eta"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.08306","kind":"arxiv","version":1},"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/2505.08306/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":"2505.08306","created_at":"2026-07-05T11:03:24.783085+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.08306v1","created_at":"2026-07-05T11:03:24.783085+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.08306","created_at":"2026-07-05T11:03:24.783085+00:00"},{"alias_kind":"pith_short_12","alias_value":"TYQTYRV6WK73","created_at":"2026-07-05T11:03:24.783085+00:00"},{"alias_kind":"pith_short_16","alias_value":"TYQTYRV6WK73YLIU","created_at":"2026-07-05T11:03:24.783085+00:00"},{"alias_kind":"pith_short_8","alias_value":"TYQTYRV6","created_at":"2026-07-05T11:03:24.783085+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11431","citing_title":"Mirror Descent Beyond Euclidean Stability: An Exponential Separation in Initialization Sensitivity","ref_index":117,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K","json":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K.json","graph_json":"https://pith.science/api/pith-number/TYQTYRV6WK73YLIUZJ4PFODT6K/graph.json","events_json":"https://pith.science/api/pith-number/TYQTYRV6WK73YLIUZJ4PFODT6K/events.json","paper":"https://pith.science/paper/TYQTYRV6"},"agent_actions":{"view_html":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K","download_json":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K.json","view_paper":"https://pith.science/paper/TYQTYRV6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.08306&json=true","fetch_graph":"https://pith.science/api/pith-number/TYQTYRV6WK73YLIUZJ4PFODT6K/graph.json","fetch_events":"https://pith.science/api/pith-number/TYQTYRV6WK73YLIUZJ4PFODT6K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K/action/storage_attestation","attest_author":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K/action/author_attestation","sign_citation":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K/action/citation_signature","submit_replication":"https://pith.science/pith/TYQTYRV6WK73YLIUZJ4PFODT6K/action/replication_record"}},"created_at":"2026-07-05T11:03:24.783085+00:00","updated_at":"2026-07-05T11:03:24.783085+00:00"}