{"id":"W4386148491","doi":"10.48550/arxiv.2308.12075","title":"Stabilizing RNN Gradients through Pre-training","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Neural Networks and Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Fonds de recherche du Québec – Nature et technologies; Natural Sciences and Engineering Research Council of Canada; Compute Canada; CHIST-ERA; Nvidia","keywords":"Initialization; Computer science; Stability (learning theory); Recurrent neural network; Exponential stability; Artificial neural network; Simple (philosophy); Differentiable function; Artificial intelligence; Field (mathematics); State space; Deep learning; Mathematical optimization; Mathematics; Machine learning; Nonlinear system","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":[0.001999358,0.001274538,0.0007813639,0.0005821033,0.0004581164,0.0009875834,0.001355247,0.001007338,0.002501009],"category_scores_gemma":[0.01569069,0.0006992609,0.000421466,0.0003839515,0.001174491,0.002509305,0.001903399,0.002219321,0.001088432],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009626879,"about_ca_system_score_gemma":0.001199049,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001830533,"about_ca_topic_score_gemma":0.004287432,"domain_scores_codex":[0.9992374,0.0002141879,0.00006274501,0.0002019628,0.0002032271,0.00008041572],"domain_scores_gemma":[0.9962699,0.001972426,0.0003941767,0.0007037726,0.000561458,0.00009829009],"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":[0.00025835,0.0001060584,0.002207004,0.0002715503,0.00006422498,0.0001345704,0.0002623648,0.7016888,0.01957568,0.04920796,0.003930437,0.222293],"study_design_scores_gemma":[0.00001056892,0.0000584422,0.000195656,0.00003064605,0.00000746782,0.00002519611,0.00001452586,0.9771718,0.00659791,0.01490408,0.0009752394,0.000008440347],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03024978,0.000348727,0.9648644,0.0002371531,0.00006661363,0.00006572746,0.00005857311,0.001692005,0.002416969],"genre_scores_gemma":[0.6189232,0.0003650012,0.3756405,0.0003171965,0.00006496796,0.0002976125,0.0002511658,0.0006288115,0.00351141],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.002501009,"threshold_uncertainty_score":0.01057374,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.214394428419426,"score_gpt":0.2296380086417515,"score_spread":0.01524358022232555,"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."}}