{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:LLWPH54QUGV3YV2K3PWRXDZ6NP","short_pith_number":"pith:LLWPH54Q","schema_version":"1.0","canonical_sha256":"5aecf3f790a1abbc574adbed1b8f3e6bfb2b13935fc3d5d8b661103b3bae6e24","source":{"kind":"arxiv","id":"2012.14368","version":2},"attestation_state":"computed","paper":{"title":"Byzantine-Resilient Non-Convex Stochastic Gradient Descent","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","cs.DS","cs.NE","math.OC"],"primary_cat":"cs.LG","authors_text":"Dan Alistarh, Faeze Ebrahimian, Jerry Li, Zeyuan Allen-Zhu","submitted_at":"2020-12-28T17:19:32Z","abstract_excerpt":"We study adversary-resilient stochastic distributed optimization, in which $m$ machines can independently compute stochastic gradients, and cooperate to jointly optimize over their local objective functions. However, an $\\alpha$-fraction of the machines are $\\textit{Byzantine}$, in that they may behave in arbitrary, adversarial ways. We consider a variant of this procedure in the challenging $\\textit{non-convex}$ case. Our main result is a new algorithm SafeguardSGD which can provably escape saddle points and find approximate local minima of the non-convex objective. The algorithm is based on "},"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":"2012.14368","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-12-28T17:19:32Z","cross_cats_sorted":["cs.DC","cs.DS","cs.NE","math.OC"],"title_canon_sha256":"a43bfbd19cc10ddeafb50830d3e4a569f6e22407dee8c7556ee4f537ed9ab337","abstract_canon_sha256":"b739ab30fa13949a8dcf19b7e5015175728e772bb8fe4933eb033b8596e9eccb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:28:44.213976Z","signature_b64":"IgLEXxLuW4kJyvaQ0GBhB+sqXyZZEhQG4WQgrf8P5lrQAeFhLbULFuFz5ya/EDnctvAa79mbzL5H68bN1xfTCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5aecf3f790a1abbc574adbed1b8f3e6bfb2b13935fc3d5d8b661103b3bae6e24","last_reissued_at":"2026-07-05T02:28:44.213477Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:28:44.213477Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Byzantine-Resilient Non-Convex Stochastic Gradient Descent","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","cs.DS","cs.NE","math.OC"],"primary_cat":"cs.LG","authors_text":"Dan Alistarh, Faeze Ebrahimian, Jerry Li, Zeyuan Allen-Zhu","submitted_at":"2020-12-28T17:19:32Z","abstract_excerpt":"We study adversary-resilient stochastic distributed optimization, in which $m$ machines can independently compute stochastic gradients, and cooperate to jointly optimize over their local objective functions. However, an $\\alpha$-fraction of the machines are $\\textit{Byzantine}$, in that they may behave in arbitrary, adversarial ways. We consider a variant of this procedure in the challenging $\\textit{non-convex}$ case. Our main result is a new algorithm SafeguardSGD which can provably escape saddle points and find approximate local minima of the non-convex objective. The algorithm is based on "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2012.14368","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/2012.14368/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":"2012.14368","created_at":"2026-07-05T02:28:44.213537+00:00"},{"alias_kind":"arxiv_version","alias_value":"2012.14368v2","created_at":"2026-07-05T02:28:44.213537+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2012.14368","created_at":"2026-07-05T02:28:44.213537+00:00"},{"alias_kind":"pith_short_12","alias_value":"LLWPH54QUGV3","created_at":"2026-07-05T02:28:44.213537+00:00"},{"alias_kind":"pith_short_16","alias_value":"LLWPH54QUGV3YV2K","created_at":"2026-07-05T02:28:44.213537+00:00"},{"alias_kind":"pith_short_8","alias_value":"LLWPH54Q","created_at":"2026-07-05T02:28:44.213537+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.07841","citing_title":"\\mathsf{VISTA}: Decentralized Machine Learning in Adversary Dominated Environments","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP","json":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP.json","graph_json":"https://pith.science/api/pith-number/LLWPH54QUGV3YV2K3PWRXDZ6NP/graph.json","events_json":"https://pith.science/api/pith-number/LLWPH54QUGV3YV2K3PWRXDZ6NP/events.json","paper":"https://pith.science/paper/LLWPH54Q"},"agent_actions":{"view_html":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP","download_json":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP.json","view_paper":"https://pith.science/paper/LLWPH54Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2012.14368&json=true","fetch_graph":"https://pith.science/api/pith-number/LLWPH54QUGV3YV2K3PWRXDZ6NP/graph.json","fetch_events":"https://pith.science/api/pith-number/LLWPH54QUGV3YV2K3PWRXDZ6NP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP/action/storage_attestation","attest_author":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP/action/author_attestation","sign_citation":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP/action/citation_signature","submit_replication":"https://pith.science/pith/LLWPH54QUGV3YV2K3PWRXDZ6NP/action/replication_record"}},"created_at":"2026-07-05T02:28:44.213537+00:00","updated_at":"2026-07-05T02:28:44.213537+00:00"}