{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:RZR3JLXWR2PQOWSLOE27RDXWBO","short_pith_number":"pith:RZR3JLXW","schema_version":"1.0","canonical_sha256":"8e63b4aef68e9f075a4b7135f88ef60b945b481500ca3960d63accf3fac90054","source":{"kind":"arxiv","id":"2005.06392","version":3},"attestation_state":"computed","paper":{"title":"On the Global Convergence Rates of Softmax Policy Gradient Methods","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Chenjun Xiao, Csaba Szepesvari, Dale Schuurmans, Jincheng Mei","submitted_at":"2020-05-13T16:01:39Z","abstract_excerpt":"We make three contributions toward better understanding policy gradient methods in the tabular setting. First, we show that with the true gradient, policy gradient with a softmax parametrization converges at a $O(1/t)$ rate, with constants depending on the problem and initialization. This result significantly expands the recent asymptotic convergence results. The analysis relies on two findings: that the softmax policy gradient satisfies a \\L{}ojasiewicz inequality, and the minimum probability of an optimal action during optimization can be bounded in terms of its initial value. Second, we ana"},"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":"2005.06392","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-05-13T16:01:39Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"d7f3ec7813de39ca0067c724066c894a63390e1d8bed048fd916c1cb3daa159c","abstract_canon_sha256":"d62c2d5bea492856efb9293b75e5c14c53ddcb0f67d5f37d276e09a05cfaba02"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:28:25.276035Z","signature_b64":"li6CafpSclq5La3r3QtUWTQ2TJUuv04HRjzGSjq7Z8vk8RY1AJwo+Kzz5Z+Iro2DT+n4VydGCAwGh4Hzh33UDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8e63b4aef68e9f075a4b7135f88ef60b945b481500ca3960d63accf3fac90054","last_reissued_at":"2026-07-05T04:28:25.275475Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:28:25.275475Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the Global Convergence Rates of Softmax Policy Gradient Methods","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Chenjun Xiao, Csaba Szepesvari, Dale Schuurmans, Jincheng Mei","submitted_at":"2020-05-13T16:01:39Z","abstract_excerpt":"We make three contributions toward better understanding policy gradient methods in the tabular setting. First, we show that with the true gradient, policy gradient with a softmax parametrization converges at a $O(1/t)$ rate, with constants depending on the problem and initialization. This result significantly expands the recent asymptotic convergence results. The analysis relies on two findings: that the softmax policy gradient satisfies a \\L{}ojasiewicz inequality, and the minimum probability of an optimal action during optimization can be bounded in terms of its initial value. Second, we ana"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2005.06392","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/2005.06392/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":"2005.06392","created_at":"2026-07-05T04:28:25.275548+00:00"},{"alias_kind":"arxiv_version","alias_value":"2005.06392v3","created_at":"2026-07-05T04:28:25.275548+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2005.06392","created_at":"2026-07-05T04:28:25.275548+00:00"},{"alias_kind":"pith_short_12","alias_value":"RZR3JLXWR2PQ","created_at":"2026-07-05T04:28:25.275548+00:00"},{"alias_kind":"pith_short_16","alias_value":"RZR3JLXWR2PQOWSL","created_at":"2026-07-05T04:28:25.275548+00:00"},{"alias_kind":"pith_short_8","alias_value":"RZR3JLXW","created_at":"2026-07-05T04:28:25.275548+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2602.01505","citing_title":"Optimal Sample Complexity for Single Time-Scale Actor-Critic with Momentum","ref_index":39,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO","json":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO.json","graph_json":"https://pith.science/api/pith-number/RZR3JLXWR2PQOWSLOE27RDXWBO/graph.json","events_json":"https://pith.science/api/pith-number/RZR3JLXWR2PQOWSLOE27RDXWBO/events.json","paper":"https://pith.science/paper/RZR3JLXW"},"agent_actions":{"view_html":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO","download_json":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO.json","view_paper":"https://pith.science/paper/RZR3JLXW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2005.06392&json=true","fetch_graph":"https://pith.science/api/pith-number/RZR3JLXWR2PQOWSLOE27RDXWBO/graph.json","fetch_events":"https://pith.science/api/pith-number/RZR3JLXWR2PQOWSLOE27RDXWBO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO/action/storage_attestation","attest_author":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO/action/author_attestation","sign_citation":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO/action/citation_signature","submit_replication":"https://pith.science/pith/RZR3JLXWR2PQOWSLOE27RDXWBO/action/replication_record"}},"created_at":"2026-07-05T04:28:25.275548+00:00","updated_at":"2026-07-05T04:28:25.275548+00:00"}