{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:QOT7MXFCOV2ROIDAT3RXDFZQRD","short_pith_number":"pith:QOT7MXFC","schema_version":"1.0","canonical_sha256":"83a7f65ca275751720609ee371973088f81ae8281c9d589d5ddb4e699cb3a34f","source":{"kind":"arxiv","id":"1909.05350","version":2},"attestation_state":"computed","paper":{"title":"The Error-Feedback Framework: Better Rates for SGD with Delayed Gradients and Compressed Communication","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Sai Praneeth Karimireddy, Sebastian U. Stich","submitted_at":"2019-09-11T20:54:49Z","abstract_excerpt":"We analyze (stochastic) gradient descent (SGD) with delayed updates on smooth quasi-convex and non-convex functions and derive concise, non-asymptotic, convergence rates. We show that the rate of convergence in all cases consists of two terms: (i) a stochastic term which is not affected by the delay, and (ii) a higher order deterministic term which is only linearly slowed down by the delay. Thus, in the presence of noise, the effects of the delay become negligible after a few iterations and the algorithm converges at the same optimal rate as standard SGD. This result extends a line of research"},"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":"1909.05350","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-09-11T20:54:49Z","cross_cats_sorted":["cs.DC","math.OC","stat.ML"],"title_canon_sha256":"77aec1e6ace6c11c019279b034d16ceb230cd1688b64a2214fba106a963869ea","abstract_canon_sha256":"183d24c34d784f8aad76efadd95753932fe9780ddac5f28fc302b59867f8d7fe"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:49:49.494362Z","signature_b64":"VYemcT3wpE2Naa3USg/tItgew2WbN6d8zGDeahY3mtVbbkf7GYozqCeb+Q3TF4mx1KvDrNy6qgn6OfN4tgyXAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"83a7f65ca275751720609ee371973088f81ae8281c9d589d5ddb4e699cb3a34f","last_reissued_at":"2026-07-05T02:49:49.493894Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:49:49.493894Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Error-Feedback Framework: Better Rates for SGD with Delayed Gradients and Compressed Communication","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Sai Praneeth Karimireddy, Sebastian U. Stich","submitted_at":"2019-09-11T20:54:49Z","abstract_excerpt":"We analyze (stochastic) gradient descent (SGD) with delayed updates on smooth quasi-convex and non-convex functions and derive concise, non-asymptotic, convergence rates. We show that the rate of convergence in all cases consists of two terms: (i) a stochastic term which is not affected by the delay, and (ii) a higher order deterministic term which is only linearly slowed down by the delay. Thus, in the presence of noise, the effects of the delay become negligible after a few iterations and the algorithm converges at the same optimal rate as standard SGD. This result extends a line of research"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1909.05350","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/1909.05350/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":"1909.05350","created_at":"2026-07-05T02:49:49.493958+00:00"},{"alias_kind":"arxiv_version","alias_value":"1909.05350v2","created_at":"2026-07-05T02:49:49.493958+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1909.05350","created_at":"2026-07-05T02:49:49.493958+00:00"},{"alias_kind":"pith_short_12","alias_value":"QOT7MXFCOV2R","created_at":"2026-07-05T02:49:49.493958+00:00"},{"alias_kind":"pith_short_16","alias_value":"QOT7MXFCOV2ROIDA","created_at":"2026-07-05T02:49:49.493958+00:00"},{"alias_kind":"pith_short_8","alias_value":"QOT7MXFC","created_at":"2026-07-05T02:49:49.493958+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.24012","citing_title":"FedSLoP: Memory-Efficient Federated Learning with Low-Rank Gradient Projection","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30634","citing_title":"One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM Pretraining","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2003.00295","citing_title":"Adaptive Federated Optimization","ref_index":236,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13434","citing_title":"Rescaled Asynchronous SGD: Optimal Distributed Optimization under Data and System Heterogeneity","ref_index":221,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25467","citing_title":"Subspace Optimization for Efficient Federated Learning under Heterogeneous Data","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24012","citing_title":"FedSLoP: Memory-Efficient Federated Learning with Low-Rank Gradient Projection","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD","json":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD.json","graph_json":"https://pith.science/api/pith-number/QOT7MXFCOV2ROIDAT3RXDFZQRD/graph.json","events_json":"https://pith.science/api/pith-number/QOT7MXFCOV2ROIDAT3RXDFZQRD/events.json","paper":"https://pith.science/paper/QOT7MXFC"},"agent_actions":{"view_html":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD","download_json":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD.json","view_paper":"https://pith.science/paper/QOT7MXFC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1909.05350&json=true","fetch_graph":"https://pith.science/api/pith-number/QOT7MXFCOV2ROIDAT3RXDFZQRD/graph.json","fetch_events":"https://pith.science/api/pith-number/QOT7MXFCOV2ROIDAT3RXDFZQRD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD/action/storage_attestation","attest_author":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD/action/author_attestation","sign_citation":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD/action/citation_signature","submit_replication":"https://pith.science/pith/QOT7MXFCOV2ROIDAT3RXDFZQRD/action/replication_record"}},"created_at":"2026-07-05T02:49:49.493958+00:00","updated_at":"2026-07-05T02:49:49.493958+00:00"}