{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:BXTHG27KXDEBXOS3EHF2PZELBI","short_pith_number":"pith:BXTHG27K","schema_version":"1.0","canonical_sha256":"0de6736beab8c81bba5b21cba7e48b0a1622a8497f89a698ced344f757c20cd1","source":{"kind":"arxiv","id":"1810.00143","version":4},"attestation_state":"computed","paper":{"title":"AdaShift: Decorrelation and Convergence of Adaptive Learning Rate Methods","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Guansong Lu, Hongwei Wang, Qingru Zhang, Weinan Zhang, Yong Yu, Zhiming Zhou","submitted_at":"2018-09-29T03:52:20Z","abstract_excerpt":"Adam is shown not being able to converge to the optimal solution in certain cases. Researchers recently propose several algorithms to avoid the issue of non-convergence of Adam, but their efficiency turns out to be unsatisfactory in practice. In this paper, we provide new insight into the non-convergence issue of Adam as well as other adaptive learning rate methods. We argue that there exists an inappropriate correlation between gradient $g_t$ and the second-moment term $v_t$ in Adam ($t$ is the timestep), which results in that a large gradient is likely to have small step size while a small g"},"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":"1810.00143","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2018-09-29T03:52:20Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"ba660c3a07969b5a3cd42f17bdbfe0548ef6ac74b35a9ae2f10d3b3442bd45cf","abstract_canon_sha256":"47411bb5f1835ba7c0c7c16a4c370026d1b1e5d894f6ec52f0d251ffa6970dba"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:42:42.985156Z","signature_b64":"ju1FOIE6YLuA9fs0NCOW4CFpjHWyjRnt7kJN+UlJCDDYFUAN5+p0sbKxiNqIQrtASv4ZAtguXJ0cQYeaEIB6CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0de6736beab8c81bba5b21cba7e48b0a1622a8497f89a698ced344f757c20cd1","last_reissued_at":"2026-05-17T23:42:42.984496Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:42:42.984496Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AdaShift: Decorrelation and Convergence of Adaptive Learning Rate Methods","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Guansong Lu, Hongwei Wang, Qingru Zhang, Weinan Zhang, Yong Yu, Zhiming Zhou","submitted_at":"2018-09-29T03:52:20Z","abstract_excerpt":"Adam is shown not being able to converge to the optimal solution in certain cases. Researchers recently propose several algorithms to avoid the issue of non-convergence of Adam, but their efficiency turns out to be unsatisfactory in practice. In this paper, we provide new insight into the non-convergence issue of Adam as well as other adaptive learning rate methods. We argue that there exists an inappropriate correlation between gradient $g_t$ and the second-moment term $v_t$ in Adam ($t$ is the timestep), which results in that a large gradient is likely to have small step size while a small g"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1810.00143","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":""},"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":"1810.00143","created_at":"2026-05-17T23:42:42.984703+00:00"},{"alias_kind":"arxiv_version","alias_value":"1810.00143v4","created_at":"2026-05-17T23:42:42.984703+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1810.00143","created_at":"2026-05-17T23:42:42.984703+00:00"},{"alias_kind":"pith_short_12","alias_value":"BXTHG27KXDEB","created_at":"2026-05-18T12:32:16.446611+00:00"},{"alias_kind":"pith_short_16","alias_value":"BXTHG27KXDEBXOS3","created_at":"2026-05-18T12:32:16.446611+00:00"},{"alias_kind":"pith_short_8","alias_value":"BXTHG27K","created_at":"2026-05-18T12:32:16.446611+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2003.00295","citing_title":"Adaptive Federated Optimization","ref_index":40,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI","json":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI.json","graph_json":"https://pith.science/api/pith-number/BXTHG27KXDEBXOS3EHF2PZELBI/graph.json","events_json":"https://pith.science/api/pith-number/BXTHG27KXDEBXOS3EHF2PZELBI/events.json","paper":"https://pith.science/paper/BXTHG27K"},"agent_actions":{"view_html":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI","download_json":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI.json","view_paper":"https://pith.science/paper/BXTHG27K","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1810.00143&json=true","fetch_graph":"https://pith.science/api/pith-number/BXTHG27KXDEBXOS3EHF2PZELBI/graph.json","fetch_events":"https://pith.science/api/pith-number/BXTHG27KXDEBXOS3EHF2PZELBI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI/action/storage_attestation","attest_author":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI/action/author_attestation","sign_citation":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI/action/citation_signature","submit_replication":"https://pith.science/pith/BXTHG27KXDEBXOS3EHF2PZELBI/action/replication_record"}},"created_at":"2026-05-17T23:42:42.984703+00:00","updated_at":"2026-05-17T23:42:42.984703+00:00"}