{"id":"W3136541527","doi":"10.1287/moor.2022.1331","title":"Convergence of Finite Memory Q Learning for POMDPs and Near Optimality of Learned Policies Under Filter Stability","year":2022,"lang":"en","type":"article","venue":"Mathematics of Operations Research","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":26,"is_retracted":false,"has_abstract":true,"ca_institutions":"Queen's University","funders":"","keywords":"Markov decision process; Partially observable Markov decision process; Convergence (economics); Bellman equation; Mathematical optimization; Mathematics; Limit (mathematics); Stability (learning theory); Filter (signal processing); Q-learning; Reinforcement learning; Optimal control; Ergodicity; Markov chain; Applied mathematics; Computer science; Markov process; Artificial intelligence; Machine learning","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003384602,0.00008664653,0.0002596189,0.0001774095,0.0004818053,0.00007525148,0.0006777839,0.00003345328,0.0001443845],"category_scores_gemma":[0.001336007,0.00008558475,0.00006018182,0.0005070648,0.000352271,0.0002130654,0.000948736,0.000302883,0.000001564468],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004904203,"about_ca_system_score_gemma":0.0002759074,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001069962,"about_ca_topic_score_gemma":0.000004428844,"domain_scores_codex":[0.9978647,0.0003533252,0.0005516199,0.0002130937,0.0007787482,0.0002385618],"domain_scores_gemma":[0.9971887,0.001353874,0.0001229222,0.0005806176,0.0007059111,0.00004800957],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000008553389,0.0001414921,0.0003556761,0.0003954954,0.0000281111,1.547482e-7,0.009657264,0.9354561,0.005158649,0.0486853,0.0000337799,0.00007942629],"study_design_scores_gemma":[0.0002377139,0.0004406867,0.0003147298,0.00001886588,0.000005949748,0.000001504789,0.00271471,0.9825591,0.01224116,0.001310095,0.00008233423,0.00007314459],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3972104,0.00003314794,0.6012613,0.0003970148,0.00002512377,0.0005194405,0.000011562,0.00001979626,0.0005222084],"genre_scores_gemma":[0.8484666,0.00001547618,0.1509396,0.000007739051,0.000004391635,0.00005756113,0.000005423431,0.00000872486,0.0004944034],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.4512562,"threshold_uncertainty_score":0.3705705,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1956737594459966,"score_gpt":0.3990674966197256,"score_spread":0.2033937371737291,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}