{"id":"W3123086775","doi":"10.1007/s10664-021-09982-4","title":"Can Offline Testing of Deep Neural Networks Replace Their Online Testing?","year":2021,"lang":"en","type":"preprint","venue":"Empirical Software Engineering","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa","funders":"H2020 European Research Council; Canada Research Chairs; Ministry of Education; National Research Foundation of Korea; Fonds National de la Recherche Luxembourg; National Research Foundation; Natural Sciences and Engineering Research Council of Canada; European Commission","keywords":"Computer science; Online and offline; Test strategy; Manual testing; Orthogonal array testing; White-box testing; Context (archaeology); Black-box testing; Exploit; Model-based testing; Machine learning; Artificial intelligence; Test case; Computer security; Operating system; Software","routes":{"ca_aff":true,"ca_fund":true,"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":[0.01404733,0.001680719,0.0008862258,0.0009884975,0.0003751558,0.001463626,0.00343587,0.002261459,0.003559345],"category_scores_gemma":[0.1136277,0.0006093204,0.0005647151,0.0006768294,0.00279619,0.006131025,0.002164724,0.003090594,0.0009798696],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001453973,"about_ca_system_score_gemma":0.001737314,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003116632,"about_ca_topic_score_gemma":0.002838701,"domain_scores_codex":[0.9864104,0.008255851,0.0006476193,0.001672725,0.002268869,0.0007444776],"domain_scores_gemma":[0.8689831,0.0976864,0.009100891,0.01731274,0.005687712,0.001229053],"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":[0.001921698,0.001324909,0.07214808,0.000623241,0.0002685792,0.0007425373,0.0003908557,0.4232561,0.01536759,0.02981069,0.008691807,0.4454539],"study_design_scores_gemma":[0.0000550876,0.0004044558,0.004140009,0.0001214575,0.00003183616,0.0001773199,0.00009124698,0.9542134,0.0113048,0.02780054,0.001626133,0.00003380236],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4172587,0.001807811,0.5588641,0.005356362,0.0005639328,0.0002164974,0.0004782394,0.004217424,0.011237],"genre_scores_gemma":[0.958581,0.000130756,0.03956475,0.0004852441,0.00005112372,0.00008724066,0.0001638474,0.0001994102,0.0007367372],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01404733,"threshold_uncertainty_score":0.07429022,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03600581497936919,"score_gpt":0.2730723730456751,"score_spread":0.2370665580663059,"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."}}