{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0002810784,0.0001509881,0.0001374107,0.0002697345,0.0001720527,0.0001955612,0.0005790874,0.00004993794,0.0002870143],"category_scores_gemma":[0.0007012482,0.0001174979,0.00009767585,0.0005473949,0.00005776765,0.0002643257,0.0001207445,0.0002936776,0.000006773283],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002690257,"about_ca_system_score_gemma":0.0002570239,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006600242,"about_ca_topic_score_gemma":0.00000133025,"domain_scores_codex":[0.9986513,0.0001613763,0.0002752971,0.0003272945,0.0004368117,0.0001479151],"domain_scores_gemma":[0.9988071,0.0001697103,0.0002544784,0.0003102837,0.0004128359,0.00004561652],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001412831,0.00002891997,0.0001174096,0.000001455745,0.00002446983,0.000002230264,0.0002033629,0.4933438,0.0002346527,0.5047289,0.00002816233,0.001272508],"study_design_scores_gemma":[0.0002346071,0.0001786447,0.0001538756,0.00006995135,0.000004145931,0.00003964693,0.00008357574,0.97425,0.0007954552,0.02404257,0.00003585607,0.0001117083],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0003690711,0.00001770929,0.9730636,0.006929736,0.0003519065,0.00009306573,0.000008929947,0.0001279238,0.01903801],"genre_scores_gemma":[0.9618095,0.00006483805,0.03717952,0.0001790445,0.0001572588,0.00001133458,0.00004529673,0.00001214791,0.0005410436],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9614404,"threshold_uncertainty_score":0.4791427,"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."}}