{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:5NXVMLVO6TVXTBDFWSM766LGNW","short_pith_number":"pith:5NXVMLVO","schema_version":"1.0","canonical_sha256":"eb6f562eaef4eb798465b499ff79666d9c1ecf1d4d110bda2f12c18863ffa3a6","source":{"kind":"arxiv","id":"2202.02831","version":3},"attestation_state":"computed","paper":{"title":"Anticorrelated Noise Injection for Improved Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","math.OC"],"primary_cat":"stat.ML","authors_text":"Antonio Orvieto, Aurelien Lucchi, Francis Bach, Frank Proske, Hans Kersting","submitted_at":"2022-02-06T18:52:21Z","abstract_excerpt":"Injecting artificial noise into gradient descent (GD) is commonly employed to improve the performance of machine learning models. Usually, uncorrelated noise is used in such perturbed gradient descent (PGD) methods. It is, however, not known if this is optimal or whether other types of noise could provide better generalization performance. In this paper, we zoom in on the problem of correlating the perturbations of consecutive PGD steps. We consider a variety of objective functions for which we find that GD with anticorrelated perturbations (\"Anti-PGD\") generalizes significantly better than GD"},"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":"2202.02831","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2022-02-06T18:52:21Z","cross_cats_sorted":["cs.LG","math.OC"],"title_canon_sha256":"c1a246dee3d99cc0b68da8a8d0252bd50c21cd5e12c7a0bb97d33bbdb3b04a4f","abstract_canon_sha256":"ee16f7c7d0a56b67d5a4f020008f73488e09907a5f9f61136296d66ab253d852"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:11:41.800404Z","signature_b64":"gDWhPuawnRWHgb5WJeY39BlxT+aIe9JU8Qc+xPfVm9FTz43XdXPfsS+qA8FgJY/735V2+b9p1xSb3X3Nf9P4CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eb6f562eaef4eb798465b499ff79666d9c1ecf1d4d110bda2f12c18863ffa3a6","last_reissued_at":"2026-07-05T06:11:41.799897Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:11:41.799897Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Anticorrelated Noise Injection for Improved Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","math.OC"],"primary_cat":"stat.ML","authors_text":"Antonio Orvieto, Aurelien Lucchi, Francis Bach, Frank Proske, Hans Kersting","submitted_at":"2022-02-06T18:52:21Z","abstract_excerpt":"Injecting artificial noise into gradient descent (GD) is commonly employed to improve the performance of machine learning models. Usually, uncorrelated noise is used in such perturbed gradient descent (PGD) methods. It is, however, not known if this is optimal or whether other types of noise could provide better generalization performance. In this paper, we zoom in on the problem of correlating the perturbations of consecutive PGD steps. We consider a variety of objective functions for which we find that GD with anticorrelated perturbations (\"Anti-PGD\") generalizes significantly better than GD"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2202.02831","kind":"arxiv","version":3},"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/2202.02831/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":"2202.02831","created_at":"2026-07-05T06:11:41.799953+00:00"},{"alias_kind":"arxiv_version","alias_value":"2202.02831v3","created_at":"2026-07-05T06:11:41.799953+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2202.02831","created_at":"2026-07-05T06:11:41.799953+00:00"},{"alias_kind":"pith_short_12","alias_value":"5NXVMLVO6TVX","created_at":"2026-07-05T06:11:41.799953+00:00"},{"alias_kind":"pith_short_16","alias_value":"5NXVMLVO6TVXTBDF","created_at":"2026-07-05T06:11:41.799953+00:00"},{"alias_kind":"pith_short_8","alias_value":"5NXVMLVO","created_at":"2026-07-05T06:11:41.799953+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.20866","citing_title":"LOSCAR-SGD: Local SGD with Communication-Computation Overlap and Delay-Corrected Sparse Model Averaging","ref_index":188,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18174","citing_title":"Ringmaster LMO: Asynchronous Linear Minimization Oracle Momentum Method","ref_index":187,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08871","citing_title":"Rennala MVR: Improved Time Complexity for Parallel Stochastic Optimization via Momentum-Based Variance Reduction","ref_index":185,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07795","citing_title":"Scalable Distributed Stochastic Optimization via Bidirectional Compression: Beyond Pessimistic Limits","ref_index":102,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW","json":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW.json","graph_json":"https://pith.science/api/pith-number/5NXVMLVO6TVXTBDFWSM766LGNW/graph.json","events_json":"https://pith.science/api/pith-number/5NXVMLVO6TVXTBDFWSM766LGNW/events.json","paper":"https://pith.science/paper/5NXVMLVO"},"agent_actions":{"view_html":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW","download_json":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW.json","view_paper":"https://pith.science/paper/5NXVMLVO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2202.02831&json=true","fetch_graph":"https://pith.science/api/pith-number/5NXVMLVO6TVXTBDFWSM766LGNW/graph.json","fetch_events":"https://pith.science/api/pith-number/5NXVMLVO6TVXTBDFWSM766LGNW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW/action/storage_attestation","attest_author":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW/action/author_attestation","sign_citation":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW/action/citation_signature","submit_replication":"https://pith.science/pith/5NXVMLVO6TVXTBDFWSM766LGNW/action/replication_record"}},"created_at":"2026-07-05T06:11:41.799953+00:00","updated_at":"2026-07-05T06:11:41.799953+00:00"}