{"id":"W4403598347","doi":"10.48550/arxiv.2409.04792","title":"Improving Deep Reinforcement Learning by Reducing the Chain Effect of Value and Policy Churn","year":2024,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Transportation and Mobility Innovations","field":"Engineering","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Alliance de recherche numérique du Canada; Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research; Fonds Québécois de la Recherche sur la Nature et les Technologies; Nvidia","keywords":"Reinforcement learning; Reinforcement; Value (mathematics); Chain (unit); Computer science; Learning effect; Artificial intelligence; Microeconomics; Economics; Machine learning; Psychology; Social psychology; Physics","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002792517,0.001151227,0.001413933,0.0004724201,0.0004646071,0.001157735,0.001517049,0.001332406,0.001827677],"category_scores_gemma":[0.01536169,0.0006878666,0.0005424874,0.000434352,0.001383008,0.002487043,0.00223467,0.00309852,0.0004787963],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001342431,"about_ca_system_score_gemma":0.002154637,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00425609,"about_ca_topic_score_gemma":0.0045285,"domain_scores_codex":[0.9987332,0.0004236881,0.00006577933,0.0002494998,0.0003634542,0.0001643313],"domain_scores_gemma":[0.9934989,0.004347073,0.0006428537,0.0006656389,0.000594758,0.0002507909],"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":[0.0001618174,0.0001686162,0.002198682,0.00008984497,0.00005267604,0.0000984439,0.0001486254,0.9112548,0.002775847,0.0123427,0.001650973,0.06905701],"study_design_scores_gemma":[0.000009279517,0.0000222496,0.0000612485,0.000005040391,0.000004070071,0.000007809347,0.000004286227,0.9961184,0.0005249379,0.003062533,0.0001774258,0.000002678145],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04451917,0.0003793773,0.9515991,0.0003272832,0.00005063688,0.00005759928,0.00003617122,0.00103422,0.001996428],"genre_scores_gemma":[0.8749444,0.0001960656,0.1221879,0.0002661852,0.00006004898,0.000116861,0.0001036712,0.0001960385,0.001928791],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.00425609,"threshold_uncertainty_score":0.01476842,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01319784069502119,"score_gpt":0.1746416961651146,"score_spread":0.1614438554700934,"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."}}