{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:HA5E3ZUGNVV3INHZY7HMZPBUQU","short_pith_number":"pith:HA5E3ZUG","schema_version":"1.0","canonical_sha256":"383a4de6866d6bb434f9c7ceccbc34852fb0a0b34704bd99441f2a2afa240c09","source":{"kind":"arxiv","id":"2405.11667","version":1},"attestation_state":"computed","paper":{"title":"The Limits and Potentials of Local SGD for Distributed Heterogeneous Learning with Intermittent Communication","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DC","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Ali Zindari, Kumar Kshitij Patel, Lingxiao Wang, Margalit Glasgow, Nathan Srebro, Nirmit Joshi, Sebastian U. Stich, Ziheng Cheng","submitted_at":"2024-05-19T20:20:03Z","abstract_excerpt":"Local SGD is a popular optimization method in distributed learning, often outperforming other algorithms in practice, including mini-batch SGD. Despite this success, theoretically proving the dominance of local SGD in settings with reasonable data heterogeneity has been difficult, creating a significant gap between theory and practice. In this paper, we provide new lower bounds for local SGD under existing first-order data heterogeneity assumptions, showing that these assumptions are insufficient to prove the effectiveness of local update steps. Furthermore, under these same assumptions, we de"},"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":"2405.11667","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-19T20:20:03Z","cross_cats_sorted":["cs.DC","math.OC","stat.ML"],"title_canon_sha256":"9aab5dd4a2ef81b7a796f37ee9f150fe3257db9178893d3a15be76eb80cac4d7","abstract_canon_sha256":"49d49c1f3c1173d6b59180832f636e9d33289a57cbc15a61de2c6a0f37fd63c2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:20:42.754343Z","signature_b64":"29QW60c1xwjZ64lIheohfmKvch8s4q/cBvVTC38WS+R+0hYZrd4JcoLyi3j9RYIGyeZ3IAk6fddstBUzwtKeCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"383a4de6866d6bb434f9c7ceccbc34852fb0a0b34704bd99441f2a2afa240c09","last_reissued_at":"2026-07-05T08:20:42.753863Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:20:42.753863Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Limits and Potentials of Local SGD for Distributed Heterogeneous Learning with Intermittent Communication","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DC","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Ali Zindari, Kumar Kshitij Patel, Lingxiao Wang, Margalit Glasgow, Nathan Srebro, Nirmit Joshi, Sebastian U. Stich, Ziheng Cheng","submitted_at":"2024-05-19T20:20:03Z","abstract_excerpt":"Local SGD is a popular optimization method in distributed learning, often outperforming other algorithms in practice, including mini-batch SGD. Despite this success, theoretically proving the dominance of local SGD in settings with reasonable data heterogeneity has been difficult, creating a significant gap between theory and practice. In this paper, we provide new lower bounds for local SGD under existing first-order data heterogeneity assumptions, showing that these assumptions are insufficient to prove the effectiveness of local update steps. Furthermore, under these same assumptions, we de"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.11667","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/2405.11667/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":"2405.11667","created_at":"2026-07-05T08:20:42.753921+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.11667v1","created_at":"2026-07-05T08:20:42.753921+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.11667","created_at":"2026-07-05T08:20:42.753921+00:00"},{"alias_kind":"pith_short_12","alias_value":"HA5E3ZUGNVV3","created_at":"2026-07-05T08:20:42.753921+00:00"},{"alias_kind":"pith_short_16","alias_value":"HA5E3ZUGNVV3INHZ","created_at":"2026-07-05T08:20:42.753921+00:00"},{"alias_kind":"pith_short_8","alias_value":"HA5E3ZUG","created_at":"2026-07-05T08:20:42.753921+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2501.14652","citing_title":"Decoupled SGDA for Games with Intermittent Strategy Communication","ref_index":35,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU","json":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU.json","graph_json":"https://pith.science/api/pith-number/HA5E3ZUGNVV3INHZY7HMZPBUQU/graph.json","events_json":"https://pith.science/api/pith-number/HA5E3ZUGNVV3INHZY7HMZPBUQU/events.json","paper":"https://pith.science/paper/HA5E3ZUG"},"agent_actions":{"view_html":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU","download_json":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU.json","view_paper":"https://pith.science/paper/HA5E3ZUG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.11667&json=true","fetch_graph":"https://pith.science/api/pith-number/HA5E3ZUGNVV3INHZY7HMZPBUQU/graph.json","fetch_events":"https://pith.science/api/pith-number/HA5E3ZUGNVV3INHZY7HMZPBUQU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU/action/storage_attestation","attest_author":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU/action/author_attestation","sign_citation":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU/action/citation_signature","submit_replication":"https://pith.science/pith/HA5E3ZUGNVV3INHZY7HMZPBUQU/action/replication_record"}},"created_at":"2026-07-05T08:20:42.753921+00:00","updated_at":"2026-07-05T08:20:42.753921+00:00"}