{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:2RZQWCEDL4I5FRV5QJ666IAJVG","short_pith_number":"pith:2RZQWCED","schema_version":"1.0","canonical_sha256":"d4730b08835f11d2c6bd827def2009a9aaaab05e5499475fbf1c2ac553b86200","source":{"kind":"arxiv","id":"2010.00406","version":4},"attestation_state":"computed","paper":{"title":"Momentum via Primal Averaging: Theoretical Insights and Learning Rate Schedules for Non-Convex Optimization","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aaron Defazio","submitted_at":"2020-10-01T13:46:32Z","abstract_excerpt":"Momentum methods are now used pervasively within the machine learning community for training non-convex models such as deep neural networks. Empirically, they out perform traditional stochastic gradient descent (SGD) approaches. In this work we develop a Lyapunov analysis of SGD with momentum (SGD+M), by utilizing a equivalent rewriting of the method known as the stochastic primal averaging (SPA) form. This analysis is much tighter than previous theory in the non-convex case, and due to this we are able to give precise insights into when SGD+M may out-perform SGD, and what hyper-parameter sche"},"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":"2010.00406","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2020-10-01T13:46:32Z","cross_cats_sorted":["math.OC","stat.ML"],"title_canon_sha256":"fc69e3cf244f357f9b0851e21224722ed59b5db699a365239db1bbf5d22f1128","abstract_canon_sha256":"61021c44ac6ac34366bb63acb5ab3a31e42ff8d8a815f68ea1ccf3f99f622a94"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:44:50.734431Z","signature_b64":"by6dfOdxk7xfP9XABVVFrAsU6Bgiu8wrCY62I8MlbtxdUoO2d/YpaaRArSJVXbut5He2OEa97Sw7+/FAvpVGDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d4730b08835f11d2c6bd827def2009a9aaaab05e5499475fbf1c2ac553b86200","last_reissued_at":"2026-07-05T02:44:50.733983Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:44:50.733983Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Momentum via Primal Averaging: Theoretical Insights and Learning Rate Schedules for Non-Convex Optimization","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aaron Defazio","submitted_at":"2020-10-01T13:46:32Z","abstract_excerpt":"Momentum methods are now used pervasively within the machine learning community for training non-convex models such as deep neural networks. Empirically, they out perform traditional stochastic gradient descent (SGD) approaches. In this work we develop a Lyapunov analysis of SGD with momentum (SGD+M), by utilizing a equivalent rewriting of the method known as the stochastic primal averaging (SPA) form. This analysis is much tighter than previous theory in the non-convex case, and due to this we are able to give precise insights into when SGD+M may out-perform SGD, and what hyper-parameter sche"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.00406","kind":"arxiv","version":4},"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/2010.00406/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":"2010.00406","created_at":"2026-07-05T02:44:50.734037+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.00406v4","created_at":"2026-07-05T02:44:50.734037+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.00406","created_at":"2026-07-05T02:44:50.734037+00:00"},{"alias_kind":"pith_short_12","alias_value":"2RZQWCEDL4I5","created_at":"2026-07-05T02:44:50.734037+00:00"},{"alias_kind":"pith_short_16","alias_value":"2RZQWCEDL4I5FRV5","created_at":"2026-07-05T02:44:50.734037+00:00"},{"alias_kind":"pith_short_8","alias_value":"2RZQWCED","created_at":"2026-07-05T02:44:50.734037+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03899","citing_title":"Denoise First, Orthogonalize Later: Understanding Momentum in Muon via Spectral Filtering","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2510.04988","citing_title":"Adaptive Memory Momentum via a Model-Based Framework for Deep Learning Optimization","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG","json":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG.json","graph_json":"https://pith.science/api/pith-number/2RZQWCEDL4I5FRV5QJ666IAJVG/graph.json","events_json":"https://pith.science/api/pith-number/2RZQWCEDL4I5FRV5QJ666IAJVG/events.json","paper":"https://pith.science/paper/2RZQWCED"},"agent_actions":{"view_html":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG","download_json":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG.json","view_paper":"https://pith.science/paper/2RZQWCED","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.00406&json=true","fetch_graph":"https://pith.science/api/pith-number/2RZQWCEDL4I5FRV5QJ666IAJVG/graph.json","fetch_events":"https://pith.science/api/pith-number/2RZQWCEDL4I5FRV5QJ666IAJVG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG/action/storage_attestation","attest_author":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG/action/author_attestation","sign_citation":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG/action/citation_signature","submit_replication":"https://pith.science/pith/2RZQWCEDL4I5FRV5QJ666IAJVG/action/replication_record"}},"created_at":"2026-07-05T02:44:50.734037+00:00","updated_at":"2026-07-05T02:44:50.734037+00:00"}