{"id":"W3128954025","doi":"10.1609/aaai.v35i9.16964","title":"Variance Penalized On-Policy and Off-Policy Actor-Critic","year":2021,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute; McGill University","funders":"","keywords":"Variance (accounting); Convergence (economics); Computer science; Reinforcement learning; Estimator; Moment (physics); Markov decision process; Expected return; Mathematical optimization; Minimum-variance unbiased estimator; Variance risk premium; Econometrics; Mathematics; Artificial intelligence; Markov process; Statistics; Economics","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0004192249,0.0002882172,0.0003397127,0.0002232419,0.0002731531,0.0005829344,0.001742806,0.0001092889,0.00006527623],"category_scores_gemma":[0.003357226,0.0002357241,0.0001121337,0.001255904,0.0003456241,0.0004480647,0.0007119945,0.0004291346,0.0001076038],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001033889,"about_ca_system_score_gemma":0.0005669993,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005235013,"about_ca_topic_score_gemma":0.000003444028,"domain_scores_codex":[0.9975829,0.00004077296,0.0005688279,0.0006314793,0.0006797189,0.0004963464],"domain_scores_gemma":[0.9978811,0.0002694826,0.0003458408,0.0005121281,0.0008358033,0.0001556483],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.00002723362,0.00006834933,0.00005984111,0.00004978302,0.00001799115,0.000002132252,0.0008564455,0.001259649,0.01703515,0.9495255,0.00008077407,0.03101717],"study_design_scores_gemma":[0.0001224709,0.0004073449,0.0004591316,0.0006618314,0.00002202047,0.00003741721,0.0004227399,0.2251096,0.5230493,0.2470201,0.002130612,0.0005575013],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1176706,0.0002723778,0.5485516,0.08803634,0.00212472,0.001789017,0.0000194822,0.0006838129,0.240852],"genre_scores_gemma":[0.9917898,0.0002425677,0.004188519,0.001398741,0.0002065033,0.00001734412,5.683319e-7,0.00001809802,0.002137908],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8741192,"threshold_uncertainty_score":0.961255,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06320478828266483,"score_gpt":0.3209269653549211,"score_spread":0.2577221770722563,"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."}}