{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FXSWNDEHOH23UEGSYLFWIXHBNV","short_pith_number":"pith:FXSWNDEH","schema_version":"1.0","canonical_sha256":"2de5668c8771f5ba10d2c2cb645ce16d7e59438f0831b97828f5b632e7d5b2e6","source":{"kind":"arxiv","id":"2401.17780","version":3},"attestation_state":"computed","paper":{"title":"A Policy Gradient Primal-Dual Algorithm for Constrained MDPs with Uniform PAC Guarantees","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Akiyoshi Sannai, Masahiro Kato, Sho Sonoda, Soichiro Nishimori, Tadashi Kozuno, Toshinori Kitamura, Wataru Kumagai, Yuki Ichihara, Yutaka Matsuo","submitted_at":"2024-01-31T12:23:24Z","abstract_excerpt":"We study a primal-dual (PD) reinforcement learning (RL) algorithm for online constrained Markov decision processes (CMDPs). Despite its widespread practical use, the existing theoretical literature on PD-RL algorithms for this problem only provides sublinear regret guarantees and fails to ensure convergence to optimal policies. In this paper, we introduce a novel policy gradient PD algorithm with uniform probably approximate correctness (Uniform-PAC) guarantees, simultaneously ensuring convergence to optimal policies, sublinear regret, and polynomial sample complexity for any target accuracy. "},"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":"2401.17780","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-01-31T12:23:24Z","cross_cats_sorted":[],"title_canon_sha256":"5f668df3418e6c36abfd1909ef7949baab9f3d5ed242ebbda248cfd396c7fb3c","abstract_canon_sha256":"4990f7d94aae90a767e6c57cb74935c0bbf99934dd9184471d7ca1b3e2a7d0f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:38:36.109375Z","signature_b64":"xpmW4xJIDya59BdM6rVSzcwNVJyolqCOyltg8hcyLNtUsk/acOCF8QXd4/2u0Vyi/3CxavBA7RjbnaN7+8LABQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2de5668c8771f5ba10d2c2cb645ce16d7e59438f0831b97828f5b632e7d5b2e6","last_reissued_at":"2026-07-05T08:38:36.108787Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:38:36.108787Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Policy Gradient Primal-Dual Algorithm for Constrained MDPs with Uniform PAC Guarantees","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Akiyoshi Sannai, Masahiro Kato, Sho Sonoda, Soichiro Nishimori, Tadashi Kozuno, Toshinori Kitamura, Wataru Kumagai, Yuki Ichihara, Yutaka Matsuo","submitted_at":"2024-01-31T12:23:24Z","abstract_excerpt":"We study a primal-dual (PD) reinforcement learning (RL) algorithm for online constrained Markov decision processes (CMDPs). Despite its widespread practical use, the existing theoretical literature on PD-RL algorithms for this problem only provides sublinear regret guarantees and fails to ensure convergence to optimal policies. In this paper, we introduce a novel policy gradient PD algorithm with uniform probably approximate correctness (Uniform-PAC) guarantees, simultaneously ensuring convergence to optimal policies, sublinear regret, and polynomial sample complexity for any target accuracy. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.17780","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/2401.17780/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":"2401.17780","created_at":"2026-07-05T08:38:36.108875+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.17780v3","created_at":"2026-07-05T08:38:36.108875+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.17780","created_at":"2026-07-05T08:38:36.108875+00:00"},{"alias_kind":"pith_short_12","alias_value":"FXSWNDEHOH23","created_at":"2026-07-05T08:38:36.108875+00:00"},{"alias_kind":"pith_short_16","alias_value":"FXSWNDEHOH23UEGS","created_at":"2026-07-05T08:38:36.108875+00:00"},{"alias_kind":"pith_short_8","alias_value":"FXSWNDEH","created_at":"2026-07-05T08:38:36.108875+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2408.16286","citing_title":"Near-Optimal Policy Identification in Robust Constrained Markov Decision Processes via Epigraph Form","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11535","citing_title":"Primal-Dual Policy Optimization for Linear CMDPs with Adversarial Losses","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV","json":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV.json","graph_json":"https://pith.science/api/pith-number/FXSWNDEHOH23UEGSYLFWIXHBNV/graph.json","events_json":"https://pith.science/api/pith-number/FXSWNDEHOH23UEGSYLFWIXHBNV/events.json","paper":"https://pith.science/paper/FXSWNDEH"},"agent_actions":{"view_html":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV","download_json":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV.json","view_paper":"https://pith.science/paper/FXSWNDEH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.17780&json=true","fetch_graph":"https://pith.science/api/pith-number/FXSWNDEHOH23UEGSYLFWIXHBNV/graph.json","fetch_events":"https://pith.science/api/pith-number/FXSWNDEHOH23UEGSYLFWIXHBNV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV/action/storage_attestation","attest_author":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV/action/author_attestation","sign_citation":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV/action/citation_signature","submit_replication":"https://pith.science/pith/FXSWNDEHOH23UEGSYLFWIXHBNV/action/replication_record"}},"created_at":"2026-07-05T08:38:36.108875+00:00","updated_at":"2026-07-05T08:38:36.108875+00:00"}