{"id":"W2750726423","doi":"","title":"Natural Value Approximators: Learning when to Trust Past Estimates","year":2017,"lang":"en","type":"article","venue":"Neural Information Processing Systems","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Observability; Reinforcement learning; Bellman equation; Classification of discontinuities; Interpolation (computer graphics); Computer science; Value (mathematics); Artificial neural network; Discontinuity (linguistics); Function approximation; Inductive bias; Artificial intelligence; Function (biology); Mathematical optimization; Algorithm; Machine learning; Mathematics; Applied mathematics; Image (mathematics); Multi-task learning","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":["sts","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.0004964991,0.000253868,0.0002660428,0.0002201867,0.001527204,0.007459884,0.001707962,0.00009125905,0.00000213696],"category_scores_gemma":[0.000508384,0.0002224242,0.00005625618,0.0001653751,0.00005705388,0.01041253,0.0004536705,0.0003775054,0.0002421855],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009867145,"about_ca_system_score_gemma":0.00006966023,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005532262,"about_ca_topic_score_gemma":3.004443e-7,"domain_scores_codex":[0.9979434,0.00004779968,0.0006552893,0.0002451419,0.0006580655,0.000450304],"domain_scores_gemma":[0.9978771,0.0000490816,0.0009045047,0.000696844,0.0003180133,0.0001544375],"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":[0.000008276865,0.000006888933,0.006560506,0.0004822203,0.00001408502,0.000003133743,0.005847658,0.9530508,0.0001633332,0.005568286,0.001403688,0.02689116],"study_design_scores_gemma":[0.0002475296,0.00006424511,0.001776781,0.0002332043,0.000005536291,0.00004514326,0.0002396231,0.9867523,0.0002528681,0.00003574179,0.01006862,0.0002784245],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01180986,0.00006900305,0.9777542,0.001042293,0.001900158,0.0005103405,7.210636e-7,0.0007575291,0.00615587],"genre_scores_gemma":[0.9807247,9.516551e-7,0.01791018,0.0003012372,0.0001797653,0.00004729583,0.00001372224,0.00001460195,0.0008075781],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9689148,"threshold_uncertainty_score":0.9997727,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01702164821844596,"score_gpt":0.264931605515689,"score_spread":0.247909957297243,"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."}}