{"id":"W2945943081","doi":"10.24963/ijcai.2019/581","title":"Metatrace Actor-Critic: Online Step-Size Tuning by Meta-gradient Descent for Reinforcement Learning Control","year":2019,"lang":"en","type":"preprint","venue":"","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Hyperparameter; Reinforcement learning; Computer science; Robustness (evolution); Gradient descent; Artificial intelligence; Nonlinear system; Function approximation; Machine learning; Mathematical optimization; Artificial neural network; Mathematics","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":["metaepi_narrow","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.001828445,0.001182264,0.001987263,0.0002731766,0.0003686469,0.00105566,0.003173666,0.0005117449,0.0002265927],"category_scores_gemma":[0.001031622,0.001031061,0.001309515,0.0002626846,0.00009783634,0.0005789765,0.002228255,0.002015257,0.00008386609],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000576345,"about_ca_system_score_gemma":0.0004270051,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000118174,"about_ca_topic_score_gemma":0.000005897225,"domain_scores_codex":[0.9931358,0.0003995803,0.001694317,0.001812058,0.001465357,0.001492918],"domain_scores_gemma":[0.9934676,0.001972348,0.001232617,0.002224004,0.0006865549,0.0004168376],"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":[0.00004008587,0.00009465939,0.000114102,0.0004536658,0.003029108,0.000005610156,0.0002824261,0.9815964,0.00044503,0.006622696,0.006021292,0.0012949],"study_design_scores_gemma":[0.001893548,0.0005246411,0.00003848562,0.0001492973,0.001708031,0.000005961144,0.0001309887,0.9231559,0.0005440687,0.00009417424,0.070669,0.00108593],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0002418253,0.001278047,0.988843,0.001628015,0.002326138,0.003341442,0.00003579024,0.0007303784,0.00157533],"genre_scores_gemma":[0.6450583,0.000319969,0.3172833,0.001955236,0.0002645992,0.0005544031,0.0003833122,0.0001577065,0.03402321],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.6715598,"threshold_uncertainty_score":0.9999813,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04272323222080878,"score_gpt":0.286499092031481,"score_spread":0.2437758598106722,"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."}}