{"id":"W4306167947","doi":"10.1109/tpami.2022.3213503","title":"Robust Losses for Learning Value Functions","year":2022,"lang":"en","type":"article","venue":"IEEE Transactions on Pattern Analysis and Machine Intelligence","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":16,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo; University of Alberta","funders":"","keywords":"Reinforcement learning; Mean squared error; Outlier; Clipping (morphology); Computer science; Bellman equation; Variance (accounting); Mathematical optimization; Robust statistics; Function (biology); Robust control; Sensitivity (control systems); Robust regression; Robustness (evolution); Mathematics; Artificial intelligence; Statistics; Control system","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":[0.0003753414,0.0001857206,0.0002445097,0.0005592271,0.001141587,0.0001730478,0.0004939105,0.00003193208,0.0003372944],"category_scores_gemma":[0.00001071026,0.0001864292,0.0002897446,0.001229829,0.00004306934,0.0001933336,0.00001664373,0.0004634327,0.0000173789],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006731146,"about_ca_system_score_gemma":0.00002812836,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003572124,"about_ca_topic_score_gemma":0.00009100027,"domain_scores_codex":[0.9984136,0.0001318203,0.0003398769,0.0005032068,0.0003448444,0.0002666619],"domain_scores_gemma":[0.9990003,0.0003010097,0.000134106,0.0004051386,0.00006741469,0.00009205852],"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":[0.000007109182,0.00005777837,0.0003791427,0.00000853697,0.0002991723,0.000002037126,0.0002144507,0.8566494,0.00003689546,0.0001854263,0.00001824903,0.1421418],"study_design_scores_gemma":[0.0000872094,0.000351781,0.0001320134,0.000004315671,0.0003348462,0.00001009412,0.0001700294,0.9938461,0.003283393,0.00006659817,0.001500598,0.0002130062],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0004171278,0.00005018211,0.9982578,0.0004422589,0.0003713194,0.0001771637,0.00002371024,0.0001437761,0.0001166315],"genre_scores_gemma":[0.9916115,0.00006854704,0.004980542,0.000355324,0.00001890448,0.0001201308,0.00001528761,0.00001459904,0.002815181],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9932773,"threshold_uncertainty_score":0.878028,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03400760162769441,"score_gpt":0.263753185754338,"score_spread":0.2297455841266436,"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."}}