{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:23TD4JQ4HOMS77WPA7NFSAUX7K","short_pith_number":"pith:23TD4JQ4","schema_version":"1.0","canonical_sha256":"d6e63e261c3b992ffecf07da590297fa821c1dfb148ed2b592b827d8e2746c32","source":{"kind":"arxiv","id":"2105.15042","version":3},"attestation_state":"computed","paper":{"title":"Semi-Anchored Multi-Step Gradient Descent Ascent Method for Structured Nonconvex-Nonconcave Composite Minimax Problems","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"math.OC","authors_text":"Donghwan Kim, Sucheol Lee","submitted_at":"2021-05-31T15:28:04Z","abstract_excerpt":"Minimax problems, such as generative adversarial network, adversarial training, and fair training, are widely solved by a multi-step gradient descent ascent (MGDA) method in practice. However, its convergence guarantee is limited. In this paper, inspired by the primal-dual hybrid gradient method, we propose a new semi-anchoring (SA) technique for the MGDA method. This makes the MGDA method find a stationary point of a structured nonconvex-nonconcave composite minimax problem; its saddle-subdifferential operator satisfies the weak Minty variational inequality condition. The resulting method, na"},"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":"2105.15042","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"math.OC","submitted_at":"2021-05-31T15:28:04Z","cross_cats_sorted":[],"title_canon_sha256":"d8d2324ebf01d173848978a5fc19d544f5453b472c8f4e5dc913c8c9c9be65d7","abstract_canon_sha256":"084b5ee3bca575d26febffd0f35a2a43754aa5c3242e51c0c3e800a6801ab66a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:30:32.001812Z","signature_b64":"r0dRNpQK+ro4BmpVz03uMAwGFXRp2MmnHH8ccYTHkKn7KKrSELwSntRDd9IDiCh1yEwHjx6qhUmy1FEJxdJbBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d6e63e261c3b992ffecf07da590297fa821c1dfb148ed2b592b827d8e2746c32","last_reissued_at":"2026-07-05T04:30:32.001389Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:30:32.001389Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Semi-Anchored Multi-Step Gradient Descent Ascent Method for Structured Nonconvex-Nonconcave Composite Minimax Problems","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"math.OC","authors_text":"Donghwan Kim, Sucheol Lee","submitted_at":"2021-05-31T15:28:04Z","abstract_excerpt":"Minimax problems, such as generative adversarial network, adversarial training, and fair training, are widely solved by a multi-step gradient descent ascent (MGDA) method in practice. However, its convergence guarantee is limited. In this paper, inspired by the primal-dual hybrid gradient method, we propose a new semi-anchoring (SA) technique for the MGDA method. This makes the MGDA method find a stationary point of a structured nonconvex-nonconcave composite minimax problem; its saddle-subdifferential operator satisfies the weak Minty variational inequality condition. The resulting method, na"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2105.15042","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/2105.15042/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":"2105.15042","created_at":"2026-07-05T04:30:32.001448+00:00"},{"alias_kind":"arxiv_version","alias_value":"2105.15042v3","created_at":"2026-07-05T04:30:32.001448+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2105.15042","created_at":"2026-07-05T04:30:32.001448+00:00"},{"alias_kind":"pith_short_12","alias_value":"23TD4JQ4HOMS","created_at":"2026-07-05T04:30:32.001448+00:00"},{"alias_kind":"pith_short_16","alias_value":"23TD4JQ4HOMS77WP","created_at":"2026-07-05T04:30:32.001448+00:00"},{"alias_kind":"pith_short_8","alias_value":"23TD4JQ4","created_at":"2026-07-05T04:30:32.001448+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.03782","citing_title":"An Improved Last-Iterate Convergence Rate for Anchored Gradient Descent Ascent","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K","json":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K.json","graph_json":"https://pith.science/api/pith-number/23TD4JQ4HOMS77WPA7NFSAUX7K/graph.json","events_json":"https://pith.science/api/pith-number/23TD4JQ4HOMS77WPA7NFSAUX7K/events.json","paper":"https://pith.science/paper/23TD4JQ4"},"agent_actions":{"view_html":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K","download_json":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K.json","view_paper":"https://pith.science/paper/23TD4JQ4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2105.15042&json=true","fetch_graph":"https://pith.science/api/pith-number/23TD4JQ4HOMS77WPA7NFSAUX7K/graph.json","fetch_events":"https://pith.science/api/pith-number/23TD4JQ4HOMS77WPA7NFSAUX7K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K/action/storage_attestation","attest_author":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K/action/author_attestation","sign_citation":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K/action/citation_signature","submit_replication":"https://pith.science/pith/23TD4JQ4HOMS77WPA7NFSAUX7K/action/replication_record"}},"created_at":"2026-07-05T04:30:32.001448+00:00","updated_at":"2026-07-05T04:30:32.001448+00:00"}