{"id":"W3186092959","doi":"10.48550/arxiv.2107.10211","title":"Differentiable Annealed Importance Sampling and the Perils of Gradient Noise","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Markov Chains and Monte Carlo Methods","field":"Mathematics","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Differentiable function; Computation; Mathematical optimization; Convergence (economics); Sublinear function; Noise (video); Mathematics; Sampling (signal processing); Computer science; Rate of convergence; Applied mathematics; Algorithm; Artificial intelligence","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":[0.0007402311,0.0002761867,0.0006941025,0.00008995063,0.0001188313,0.00004886276,0.0003819063,0.0001970075,0.00004004758],"category_scores_gemma":[0.0003072858,0.0002257059,0.0003112332,0.0001947877,0.0002739841,0.00005402487,0.0008818494,0.0004116004,3.353326e-8],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006536802,"about_ca_system_score_gemma":0.00007675649,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001362932,"about_ca_topic_score_gemma":0.0001572703,"domain_scores_codex":[0.9984533,0.0003039271,0.000310621,0.0005958411,0.00008609296,0.0002501986],"domain_scores_gemma":[0.9977086,0.0006303514,0.0004289411,0.0009402578,0.0001870649,0.0001048121],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0007807088,0.0004795519,0.03785856,0.003811794,0.001186215,0.0002862908,0.005797175,0.006494552,0.0007330512,0.9415472,0.0003487746,0.0006761313],"study_design_scores_gemma":[0.01597529,0.0002355417,0.01049767,0.003417258,0.004903265,0.00006472396,0.01881403,0.3550027,0.0033067,0.5826491,0.001659709,0.00347397],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9416779,0.0005377775,0.05516871,0.00003906738,0.0002091104,0.0004045521,0.00002430503,0.00003709183,0.001901526],"genre_scores_gemma":[0.9918517,0.0009347355,0.005408688,0.00002694796,0.00004832149,0.00000236163,0.00001318876,0.0000263598,0.001687734],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.358898,"threshold_uncertainty_score":0.9204021,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1663564451730863,"score_gpt":0.2479423587077539,"score_spread":0.0815859135346676,"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."}}