{"id":"W2771330107","doi":"10.1162/neco_a_01111","title":"An Empirical Evaluation of Rule Extraction from Recurrent Neural Networks","year":2018,"lang":"en","type":"article","venue":"Neural Computation","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":58,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Recurrent neural network; Computer science; Artificial intelligence; Black box; Machine learning; Artificial neural network; Rule-based machine translation","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":[0.0004972932,0.0001277871,0.0001347835,0.0001098494,0.0001115077,0.0001300365,0.0003999668,0.00009287206,0.00001514609],"category_scores_gemma":[0.00007342139,0.0001174202,0.00004325244,0.000387095,0.00005482517,0.001113339,0.0000722659,0.0001869937,0.000004384508],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007659342,"about_ca_system_score_gemma":0.00003443721,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005908854,"about_ca_topic_score_gemma":0.00001375796,"domain_scores_codex":[0.9981959,0.0003678486,0.0002958047,0.0003838306,0.0005942116,0.0001624277],"domain_scores_gemma":[0.9986442,0.00009560632,0.000248567,0.0002814967,0.0006694161,0.00006070178],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00002634612,0.00008299962,0.0008987187,0.000004333192,0.000005316977,0.000001488805,0.0004895927,0.01008575,0.006583991,0.0001639364,0.0002855527,0.981372],"study_design_scores_gemma":[0.0002126826,0.0002688811,0.00874653,0.00001470973,0.00001772279,0.000006723925,0.000008249859,0.9693531,0.008847818,0.01239558,0.000009191734,0.0001187694],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4513662,0.0002271155,0.547312,0.0001520045,0.0005366175,0.000133768,0.000001155593,0.0002430396,0.00002816072],"genre_scores_gemma":[0.8731139,0.000001169533,0.1263552,0.0001485498,0.0003118059,0.000008678036,0.00005162516,0.000008109295,0.000001001161],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9812532,"threshold_uncertainty_score":0.4788255,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05760352501565048,"score_gpt":0.4038145947743336,"score_spread":0.3462110697586831,"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."}}