{"id":"W3167824561","doi":"","title":"Beyond Variance Reduction: Understanding the True Impact of Baselines on Policy Optimization","year":2021,"lang":"en","type":"article","venue":"International Conference on Machine Learning","topic":"Stochastic Gradient Optimization Techniques","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Reinforcement learning; Mathematical optimization; Variance reduction; Computer science; Variance (accounting); Noise (video); Optimization problem; Baseline (sea); Focus (optics); Stochastic optimization; Curvature; Mathematics; Artificial intelligence; Image (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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01499594,0.001191166,0.002373694,0.001207666,0.001213793,0.004724477,0.001726469,0.002665558,0.003132934],"category_scores_gemma":[0.1018435,0.0007692768,0.0009015687,0.0009444395,0.003271923,0.009420582,0.003334367,0.005025049,0.0005655211],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002425235,"about_ca_system_score_gemma":0.002137601,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002958592,"about_ca_topic_score_gemma":0.00213221,"domain_scores_codex":[0.9935628,0.003409969,0.0002541495,0.001169645,0.001076774,0.0005265975],"domain_scores_gemma":[0.9578428,0.03219952,0.002767263,0.004788973,0.001470569,0.0009308761],"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":[0.0004572823,0.0002386873,0.005536872,0.0002380159,0.0001897479,0.0001515996,0.0003385766,0.5708395,0.002209545,0.3543149,0.002726644,0.06275854],"study_design_scores_gemma":[0.0000276524,0.0001784089,0.001192299,0.00004878941,0.00002230438,0.00003926572,0.00005705539,0.7857615,0.0008446857,0.2108904,0.0009128501,0.00002474494],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1464367,0.002662103,0.8273761,0.005147487,0.000206043,0.0001461536,0.0002582837,0.001008748,0.01675837],"genre_scores_gemma":[0.9494423,0.0005534137,0.04729109,0.0005353694,0.0001014918,0.0001183442,0.0001387403,0.0002593548,0.001559832],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01499594,"threshold_uncertainty_score":0.07930702,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04983852392367744,"score_gpt":0.3330668308248156,"score_spread":0.2832283069011381,"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."}}