{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:WILZIHPYROQM6YGM2CACMDVHKT","short_pith_number":"pith:WILZIHPY","schema_version":"1.0","canonical_sha256":"b217941df88ba0cf60ccd080260ea754cd10f2c4d5234e786ad95b231414ab21","source":{"kind":"arxiv","id":"1705.10412","version":2},"attestation_state":"computed","paper":{"title":"Gradient Descent Can Take Exponential Time to Escape Saddle Points","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"math.OC","authors_text":"Aarti Singh, Barnabas Poczos, Chi Jin, Jason D. Lee, Michael I. Jordan, Simon S. Du","submitted_at":"2017-05-29T23:03:01Z","abstract_excerpt":"Although gradient descent (GD) almost always escapes saddle points asymptotically [Lee et al., 2016], this paper shows that even with fairly natural random initialization schemes and non-pathological functions, GD can be significantly slowed down by saddle points, taking exponential time to escape. On the other hand, gradient descent with perturbations [Ge et al., 2015, Jin et al., 2017] is not slowed down by saddle points - it can find an approximate local minimizer in polynomial time. This result implies that GD is inherently slower than perturbed GD, and justifies the importance of adding p"},"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":"1705.10412","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"math.OC","submitted_at":"2017-05-29T23:03:01Z","cross_cats_sorted":["cs.LG","stat.ML"],"title_canon_sha256":"37a31bb45fa84b75f2872147e9fe69aac38512d29dee40aafc79995254da9d1f","abstract_canon_sha256":"9572245b2280015d0a1e299a2c3228c935fd5d7baba638367c74b249bbf1e26d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:31:19.318785Z","signature_b64":"Ds+wK7N+pkhlJUZiNbLMjceBALuCKexLESFBt2uLZvhuTRZtgvat0AZT+e6K3FOxiUmLoaoS412JXD2+dbo3Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b217941df88ba0cf60ccd080260ea754cd10f2c4d5234e786ad95b231414ab21","last_reissued_at":"2026-05-18T00:31:19.318059Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:31:19.318059Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Gradient Descent Can Take Exponential Time to Escape Saddle Points","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"math.OC","authors_text":"Aarti Singh, Barnabas Poczos, Chi Jin, Jason D. Lee, Michael I. Jordan, Simon S. Du","submitted_at":"2017-05-29T23:03:01Z","abstract_excerpt":"Although gradient descent (GD) almost always escapes saddle points asymptotically [Lee et al., 2016], this paper shows that even with fairly natural random initialization schemes and non-pathological functions, GD can be significantly slowed down by saddle points, taking exponential time to escape. On the other hand, gradient descent with perturbations [Ge et al., 2015, Jin et al., 2017] is not slowed down by saddle points - it can find an approximate local minimizer in polynomial time. This result implies that GD is inherently slower than perturbed GD, and justifies the importance of adding p"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1705.10412","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":""},"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":"1705.10412","created_at":"2026-05-18T00:31:19.318167+00:00"},{"alias_kind":"arxiv_version","alias_value":"1705.10412v2","created_at":"2026-05-18T00:31:19.318167+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1705.10412","created_at":"2026-05-18T00:31:19.318167+00:00"},{"alias_kind":"pith_short_12","alias_value":"WILZIHPYROQM","created_at":"2026-05-18T12:31:53.515858+00:00"},{"alias_kind":"pith_short_16","alias_value":"WILZIHPYROQM6YGM","created_at":"2026-05-18T12:31:53.515858+00:00"},{"alias_kind":"pith_short_8","alias_value":"WILZIHPY","created_at":"2026-05-18T12:31:53.515858+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"1907.01849","citing_title":"Distributed Learning in Non-Convex Environments -- Part II: Polynomial Escape from Saddle-Points","ref_index":23,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT","json":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT.json","graph_json":"https://pith.science/api/pith-number/WILZIHPYROQM6YGM2CACMDVHKT/graph.json","events_json":"https://pith.science/api/pith-number/WILZIHPYROQM6YGM2CACMDVHKT/events.json","paper":"https://pith.science/paper/WILZIHPY"},"agent_actions":{"view_html":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT","download_json":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT.json","view_paper":"https://pith.science/paper/WILZIHPY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1705.10412&json=true","fetch_graph":"https://pith.science/api/pith-number/WILZIHPYROQM6YGM2CACMDVHKT/graph.json","fetch_events":"https://pith.science/api/pith-number/WILZIHPYROQM6YGM2CACMDVHKT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT/action/storage_attestation","attest_author":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT/action/author_attestation","sign_citation":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT/action/citation_signature","submit_replication":"https://pith.science/pith/WILZIHPYROQM6YGM2CACMDVHKT/action/replication_record"}},"created_at":"2026-05-18T00:31:19.318167+00:00","updated_at":"2026-05-18T00:31:19.318167+00:00"}