{"id":"https://openalex.org/W3207833002","doi":"https://doi.org/10.1109/tac.2025.3540965","title":"Beyond Exact Gradients: Convergence of Stochastic Soft-Max Policy Gradient Methods With Entropy Regularization","display_name":"Beyond Exact Gradients: Convergence of Stochastic Soft-Max Policy Gradient Methods With Entropy Regularization","publication_year":2025,"publication_date":"2025-02-11","ids":{"openalex":"https://openalex.org/W3207833002","doi":"https://doi.org/10.1109/tac.2025.3540965","mag":"3207833002"},"language":"en","primary_location":{"id":"doi:10.1109/tac.2025.3540965","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tac.2025.3540965","pdf_url":null,"source":{"id":"https://openalex.org/S184954342","display_name":"IEEE Transactions on Automatic Control","issn_l":"0018-9286","issn":["0018-9286","1558-2523","2334-3303"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Automatic Control","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"http://export.arxiv.org/pdf/2110.10117","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102879163","display_name":"Yuhao Ding","orcid":"https://orcid.org/0000-0002-1521-7566"},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuhao Ding","raw_affiliation_strings":["Department of Industrial Engineering and Operations Research, University of California, Berkeley, Berkeley, CA, USA","Department of Industrial Engineering and Operations Research, University of California, Berkeley, CA, USA"],"raw_orcid":"https://orcid.org/0000-0002-1521-7566","affiliations":[{"raw_affiliation_string":"Department of Industrial Engineering and Operations Research, University of California, Berkeley, Berkeley, CA, USA","institution_ids":["https://openalex.org/I95457486"]},{"raw_affiliation_string":"Department of Industrial Engineering and Operations Research, University of California, Berkeley, CA, USA","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089647297","display_name":"Junzi Zhang","orcid":"https://orcid.org/0000-0002-5086-0063"},"institutions":[{"id":"https://openalex.org/I91036609","display_name":"Citadel","ror":"https://ror.org/01vwr6t80","country_code":"US","type":"education","lineage":["https://openalex.org/I91036609"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Junzi Zhang","raw_affiliation_strings":["Citadel Securities, Chicago, IL, USA","Citadel Securities (work done prior to joining Citadel Securities), Chicago, IL, USA"],"raw_orcid":"https://orcid.org/0000-0002-5086-0063","affiliations":[{"raw_affiliation_string":"Citadel Securities, Chicago, IL, USA","institution_ids":["https://openalex.org/I91036609"]},{"raw_affiliation_string":"Citadel Securities (work done prior to joining Citadel Securities), Chicago, IL, USA","institution_ids":["https://openalex.org/I91036609"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5069203447","display_name":"Hyunin Lee","orcid":"https://orcid.org/0000-0003-1718-9594"},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hyunin Lee","raw_affiliation_strings":["Department of Mechanical Engineering, University of California, Berkeley, Berkeley, CA, USA","Department of Mechanical Engineering, University of California, Berkeley, CA, USA"],"raw_orcid":"https://orcid.org/0000-0003-1718-9594","affiliations":[{"raw_affiliation_string":"Department of Mechanical Engineering, University of California, Berkeley, Berkeley, CA, USA","institution_ids":["https://openalex.org/I95457486"]},{"raw_affiliation_string":"Department of Mechanical Engineering, University of California, Berkeley, CA, USA","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5001393548","display_name":"Javad Lavaei","orcid":"https://orcid.org/0000-0003-4294-1338"},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Javad Lavaei","raw_affiliation_strings":["Department of Industrial Engineering and Operations Research, University of California, Berkeley, Berkeley, CA, USA","Department of Industrial Engineering and Operations Research, University of California, Berkeley, CA, USA"],"raw_orcid":"https://orcid.org/0000-0003-4294-1338","affiliations":[{"raw_affiliation_string":"Department of Industrial Engineering and Operations Research, University of California, Berkeley, Berkeley, CA, USA","institution_ids":["https://openalex.org/I95457486"]},{"raw_affiliation_string":"Department of Industrial Engineering and Operations Research, University of California, Berkeley, CA, USA","institution_ids":["https://openalex.org/I95457486"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.396,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.81857899,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":"70","issue":"8","first_page":"5129","last_page":"5144"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9997000098228455,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9997000098228455,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12101","display_name":"Advanced Bandit Algorithms Research","score":0.9988999962806702,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9926999807357788,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/estimator","display_name":"Estimator","score":0.6556838154792786},{"id":"https://openalex.org/keywords/entropy","display_name":"Entropy (arrow of time)","score":0.582999587059021},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.5758589506149292},{"id":"https://openalex.org/keywords/mathematical-optimization","display_name":"Mathematical optimization","score":0.5591661334037781},{"id":"https://openalex.org/keywords/applied-mathematics","display_name":"Applied mathematics","score":0.5251426696777344},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.4627886116504669},{"id":"https://openalex.org/keywords/logarithm","display_name":"Logarithm","score":0.4418162405490875},{"id":"https://openalex.org/keywords/initialization","display_name":"Initialization","score":0.42358237504959106},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.311855673789978},{"id":"https://openalex.org/keywords/mathematical-analysis","display_name":"Mathematical analysis","score":0.14380159974098206},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.10461625456809998},{"id":"https://openalex.org/keywords/statistics","display_name":"Statistics","score":0.09969088435173035},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.0685063898563385}],"concepts":[{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.6556838154792786},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.582999587059021},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5758589506149292},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.5591661334037781},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.5251426696777344},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.4627886116504669},{"id":"https://openalex.org/C39927690","wikidata":"https://www.wikidata.org/wiki/Q11197","display_name":"Logarithm","level":2,"score":0.4418162405490875},{"id":"https://openalex.org/C114466953","wikidata":"https://www.wikidata.org/wiki/Q6034165","display_name":"Initialization","level":2,"score":0.42358237504959106},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.311855673789978},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.14380159974098206},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.10461625456809998},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.09969088435173035},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0685063898563385},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/tac.2025.3540965","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tac.2025.3540965","pdf_url":null,"source":{"id":"https://openalex.org/S184954342","display_name":"IEEE Transactions on Automatic Control","issn_l":"0018-9286","issn":["0018-9286","1558-2523","2334-3303"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Automatic Control","raw_type":"journal-article"},{"id":"mag:3207833002","is_oa":true,"landing_page_url":"http://export.arxiv.org/pdf/2110.10117","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null}],"best_oa_location":{"id":"mag:3207833002","is_oa":true,"landing_page_url":"http://export.arxiv.org/pdf/2110.10117","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3014366370","display_name":null,"funder_award_id":"N000142412673","funder_id":"https://openalex.org/F4320337345","funder_display_name":"Office of Naval Research"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320337345","display_name":"Office of Naval Research","ror":"https://ror.org/00rk2pe57"},{"id":"https://openalex.org/F4320338279","display_name":"Air Force Office of Scientific Research","ror":"https://ror.org/011e9bt93"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":75,"referenced_works":["https://openalex.org/W6908809","https://openalex.org/W64088143","https://openalex.org/W594357522","https://openalex.org/W1491706803","https://openalex.org/W1971942712","https://openalex.org/W1991667331","https://openalex.org/W1993411524","https://openalex.org/W2099111195","https://openalex.org/W2119717200","https://openalex.org/W2121863487","https://openalex.org/W2144446635","https://openalex.org/W2146502635","https://openalex.org/W2155027007","https://openalex.org/W2235056388","https://openalex.org/W2554120691","https://openalex.org/W2609650878","https://openalex.org/W2619268125","https://openalex.org/W2783096200","https://openalex.org/W2948432982","https://openalex.org/W2962821147","https://openalex.org/W2962902376","https://openalex.org/W2964043796","https://openalex.org/W2969951231","https://openalex.org/W2978644431","https://openalex.org/W2998050631","https://openalex.org/W3012939900","https://openalex.org/W3034426742","https://openalex.org/W3036276779","https://openalex.org/W3041970508","https://openalex.org/W3093528669","https://openalex.org/W3103182070","https://openalex.org/W3109546547","https://openalex.org/W3127686539","https://openalex.org/W3133371913","https://openalex.org/W3135724079","https://openalex.org/W3169411945","https://openalex.org/W3171210634","https://openalex.org/W3186110329","https://openalex.org/W3217314940","https://openalex.org/W4214717370","https://openalex.org/W4221157223","https://openalex.org/W4233840023","https://openalex.org/W4255900881","https://openalex.org/W4286900472","https://openalex.org/W6600284362","https://openalex.org/W6631190155","https://openalex.org/W6681435938","https://openalex.org/W6683204974","https://openalex.org/W6692846177","https://openalex.org/W6734206676","https://openalex.org/W6734517396","https://openalex.org/W6736495777","https://openalex.org/W6747473740","https://openalex.org/W6756287877","https://openalex.org/W6767052781","https://openalex.org/W6769196924","https://openalex.org/W6775255735","https://openalex.org/W6776824048","https://openalex.org/W6779228417","https://openalex.org/W6780386840","https://openalex.org/W6780486084","https://openalex.org/W6782593158","https://openalex.org/W6784698908","https://openalex.org/W6786084949","https://openalex.org/W6791168167","https://openalex.org/W6791509694","https://openalex.org/W6791665192","https://openalex.org/W6794588271","https://openalex.org/W6797352476","https://openalex.org/W6798832068","https://openalex.org/W6802547569","https://openalex.org/W6803111450","https://openalex.org/W6809872370","https://openalex.org/W6840151816","https://openalex.org/W7055423279"],"related_works":["https://openalex.org/W3204184292","https://openalex.org/W3176564347","https://openalex.org/W2355833770","https://openalex.org/W1985458517","https://openalex.org/W3031039437","https://openalex.org/W183202219","https://openalex.org/W3095877357","https://openalex.org/W2072565696","https://openalex.org/W2050451745","https://openalex.org/W2378903222"],"abstract_inverted_index":{"Entropy":[0],"regularization":[1],"is":[2,207],"an":[3,87],"efficient":[4],"technique":[5],"for":[6,201,216],"encouraging":[7],"exploration":[8],"and":[9,92,164,190,212],"preventing":[10],"a":[11,97,136,143,166,185,191],"premature":[12],"convergence":[13,51,188,211],"of":[14,28,74,155,194],"(vanilla)":[15],"policy":[16,48,121],"gradient":[17,62],"(PG)":[18],"methods":[19,44],"in":[20,114,147,170],"reinforcement":[21],"learning":[22],"(RL).":[23],"However,":[24],"the":[25,40,46,71,93,109,118,125,128,148,153,156,161,171,176,180,202,208,217],"theoretical":[26],"understanding":[27],"entropy-regularized":[29,42,219],"RL":[30],"algorithms":[31],"has":[32,52],"been":[33,56],"limited.":[34],"In":[35],"this":[36,67],"article,":[37],"we":[38,69],"revisit":[39],"classical":[41],"PG":[43,78,139,221],"with":[45,80,84],"soft-max":[47],"parametrization,":[49],"whose":[50],"so":[53],"far":[54],"only":[55],"established":[57],"assuming":[58],"access":[59],"to":[60,117,151,160],"exact":[61],"oracles.":[63],"To":[64],"go":[65],"beyond":[66],"scenario,":[68],"propose":[70,135],"first":[72,149,209],"set":[73],"(nearly)":[75],"unbiased":[76,88,99],"stochastic":[77,138,157,218],"estimators":[79,110],"trajectory-level":[81],"entropy":[82,126],"regularization,":[83],"one":[85,95],"being":[86,96],"visitation":[89],"measure-based":[90],"estimator":[91],"other":[94],"nearly":[98],"yet":[100],"more":[101],"practical":[102],"trajectory-based":[103],"estimator.":[104],"We":[105,133,183],"prove":[106],"that":[107,141],"although":[108],"themselves":[111],"are":[112,130],"unbounded":[113],"general":[115],"due":[116,159],"additional":[119],"logarithmic":[120],"rewards":[122],"introduced":[123],"by":[124,174],"term,":[127],"variances":[129],"uniformly":[131],"bounded.":[132],"then":[134],"two-phase":[137],"algorithm":[140],"uses":[142,165],"large":[144],"batch":[145,168],"size":[146,169],"phase":[150,173],"overcome":[152],"challenge":[154],"approximation":[158],"noncoercive":[162],"landscape,":[163],"small":[167],"second":[172],"leveraging":[175],"curvature":[177],"information":[178],"around":[179],"optimal":[181],"policy.":[182],"establish":[184],"global":[186,210],"optimality":[187],"result":[189,206],"sample":[192,213],"complexity":[193,214],"<inline-formula":[195],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[196],"xmlns:xlink=\"http://www.w3.org/1999/xlink\"><tex-math":[197],"notation=\"LaTeX\">$\\widetilde{\\mathcal":[198],"{O}}(\\frac{1}{\\epsilon":[199],"^{2}})$</tex-math></inline-formula>":[200],"proposed":[203],"algorithm.":[204],"Our":[205],"results":[215],"vanilla":[220],"method.":[222]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2025-12-22T23:10:17.713674","created_date":"2025-10-10T00:00:00"}