{"id":"W2996690709","doi":"10.1109/aicas48895.2020.9073854","title":"Layerwise Noise Maximisation to Train Low-Energy Deep Neural Networks","year":2020,"lang":"en","type":"preprint","venue":"","topic":"Advanced Neural Network Applications","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"Polytechnique Montréal","funders":"Natural Sciences and Engineering Research Council of Canada; Institut de Valorisation des Données","keywords":"Overhead (engineering); Reliability (semiconductor); Computer science; Artificial neural network; Energy (signal processing); Stochastic gradient descent; Inference; Energy consumption; Deep neural networks; Noise (video); Exponential function; Layer (electronics); Algorithm; Artificial intelligence; Statistics; Engineering; Mathematics","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":[0.001308098,0.00150405,0.001023111,0.000511468,0.0003060172,0.0008033702,0.001698143,0.001400068,0.003002489],"category_scores_gemma":[0.005511574,0.0008300684,0.0006902789,0.0007197562,0.0008465084,0.001343934,0.001271615,0.002176634,0.001037638],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001037343,"about_ca_system_score_gemma":0.00109416,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002707348,"about_ca_topic_score_gemma":0.004482623,"domain_scores_codex":[0.9995137,0.0001468156,0.00003127328,0.0001059394,0.0001422534,0.00005998731],"domain_scores_gemma":[0.9988596,0.0007144063,0.00008747039,0.0001065062,0.000201804,0.00003012777],"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":[0.00009864516,0.00003531253,0.0002507162,0.0000826791,0.00004579763,0.0000331696,0.00003379725,0.92215,0.003978095,0.004797494,0.001432208,0.06706211],"study_design_scores_gemma":[0.000004222969,0.00001212504,0.00003322094,0.000004834561,0.000003912191,0.000004331805,0.0000017965,0.9960114,0.001338665,0.002357636,0.000225712,0.000002132892],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.009723476,0.0002904474,0.9877188,0.0001156555,0.00003439869,0.00002821916,0.00005972596,0.001143327,0.0008859638],"genre_scores_gemma":[0.529338,0.000424366,0.4621114,0.0003206605,0.0001131289,0.0003816157,0.0005763905,0.0006514113,0.006083019],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.003002489,"threshold_uncertainty_score":0.01004434,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02502243469470757,"score_gpt":0.2590967731519528,"score_spread":0.2340743384572453,"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."}}