{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:JPF3RTT6JJN3PETA7DJFSPTX7F","short_pith_number":"pith:JPF3RTT6","schema_version":"1.0","canonical_sha256":"4bcbb8ce7e4a5bb79260f8d2593e77f978f71acbad4d62b5adc552f77cbda95d","source":{"kind":"arxiv","id":"2410.15637","version":2},"attestation_state":"computed","paper":{"title":"Large Deviation Upper Bounds and Improved MSE Rates of Nonlinear SGD: Heavy-tailed Noise and Power of Symmetry","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","math.PR"],"primary_cat":"cs.LG","authors_text":"Aleksandar Armacki, Dragana Bajovic, Dusan Jakovetic, Shuhua Yu, Soummya Kar","submitted_at":"2024-10-21T04:50:57Z","abstract_excerpt":"We study large deviation upper bounds and mean-squared error (MSE) guarantees of a general framework of nonlinear stochastic gradient methods in the online setting, in the presence of heavy-tailed noise. Unlike existing works that rely on the closed form of a nonlinearity (typically clipping), our framework treats the nonlinearity in a black-box manner, allowing us to provide unified guarantees for a broad class of bounded nonlinearities, including many popular ones, like sign, quantization, normalization, as well as component-wise and joint clipping. We provide several strong results for a br"},"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":"2410.15637","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-21T04:50:57Z","cross_cats_sorted":["math.OC","math.PR"],"title_canon_sha256":"5807867a639662e77d0c4fc516caba8cc883d69a7938892d3d729fd34c15ba4e","abstract_canon_sha256":"027cef568aedc74ef9bf6dd170e109b7bb2be24857b64ed98a443bb42109f6e5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:37:35.467189Z","signature_b64":"h8Orcx9yJa/03BSBhNx3HegWJUPd9UEJ9UlAJGOYFXlXwZBi+IYmO10ieh0Slj+Ee5OJ3EPAkJlJZIAWqSeLCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4bcbb8ce7e4a5bb79260f8d2593e77f978f71acbad4d62b5adc552f77cbda95d","last_reissued_at":"2026-07-05T10:37:35.466603Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:37:35.466603Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Large Deviation Upper Bounds and Improved MSE Rates of Nonlinear SGD: Heavy-tailed Noise and Power of Symmetry","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","math.PR"],"primary_cat":"cs.LG","authors_text":"Aleksandar Armacki, Dragana Bajovic, Dusan Jakovetic, Shuhua Yu, Soummya Kar","submitted_at":"2024-10-21T04:50:57Z","abstract_excerpt":"We study large deviation upper bounds and mean-squared error (MSE) guarantees of a general framework of nonlinear stochastic gradient methods in the online setting, in the presence of heavy-tailed noise. Unlike existing works that rely on the closed form of a nonlinearity (typically clipping), our framework treats the nonlinearity in a black-box manner, allowing us to provide unified guarantees for a broad class of bounded nonlinearities, including many popular ones, like sign, quantization, normalization, as well as component-wise and joint clipping. We provide several strong results for a br"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.15637","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/2410.15637/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":"2410.15637","created_at":"2026-07-05T10:37:35.466669+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.15637v2","created_at":"2026-07-05T10:37:35.466669+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.15637","created_at":"2026-07-05T10:37:35.466669+00:00"},{"alias_kind":"pith_short_12","alias_value":"JPF3RTT6JJN3","created_at":"2026-07-05T10:37:35.466669+00:00"},{"alias_kind":"pith_short_16","alias_value":"JPF3RTT6JJN3PETA","created_at":"2026-07-05T10:37:35.466669+00:00"},{"alias_kind":"pith_short_8","alias_value":"JPF3RTT6","created_at":"2026-07-05T10:37:35.466669+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2505.03736","citing_title":"Decentralized Nonconvex Optimization under Heavy-Tailed Noise: Normalization and Optimal Convergence","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2510.06141","citing_title":"High-Probability Convergence Guarantees of Decentralized SGD","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2510.06141","citing_title":"High-Probability Convergence Guarantees of Decentralized SGD","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F","json":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F.json","graph_json":"https://pith.science/api/pith-number/JPF3RTT6JJN3PETA7DJFSPTX7F/graph.json","events_json":"https://pith.science/api/pith-number/JPF3RTT6JJN3PETA7DJFSPTX7F/events.json","paper":"https://pith.science/paper/JPF3RTT6"},"agent_actions":{"view_html":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F","download_json":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F.json","view_paper":"https://pith.science/paper/JPF3RTT6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.15637&json=true","fetch_graph":"https://pith.science/api/pith-number/JPF3RTT6JJN3PETA7DJFSPTX7F/graph.json","fetch_events":"https://pith.science/api/pith-number/JPF3RTT6JJN3PETA7DJFSPTX7F/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F/action/storage_attestation","attest_author":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F/action/author_attestation","sign_citation":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F/action/citation_signature","submit_replication":"https://pith.science/pith/JPF3RTT6JJN3PETA7DJFSPTX7F/action/replication_record"}},"created_at":"2026-07-05T10:37:35.466669+00:00","updated_at":"2026-07-05T10:37:35.466669+00:00"}