{"id":"W7126423241","doi":"10.21428/594757db.269457de","title":"GCE: Confidence Calibration Error for ImprovedTrustworthiness of Graph Neural Networks","year":2024,"lang":"en","type":"article","venue":"","topic":"Advanced Graph Neural Networks","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; Toronto Metropolitan University","funders":"","keywords":"Graph; Artificial neural network; Confidence interval; Calibration; Homogeneous; Pattern recognition (psychology); Code (set theory); Training set; Error function","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006116732,0.001534806,0.001221725,0.001597611,0.000782612,0.001748083,0.003223498,0.002467779,0.00281656],"category_scores_gemma":[0.05280172,0.0006303985,0.0007435005,0.0009660443,0.001670275,0.004182784,0.003273112,0.003798968,0.0007222724],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002527152,"about_ca_system_score_gemma":0.001535739,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006186683,"about_ca_topic_score_gemma":0.00736019,"domain_scores_codex":[0.9975177,0.0007954689,0.0001618896,0.0005656374,0.000748677,0.000210676],"domain_scores_gemma":[0.9820918,0.009678373,0.001828599,0.003192739,0.002712556,0.0004959711],"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":[0.0003197816,0.0001352949,0.007226652,0.0001925565,0.0001297661,0.0001563653,0.0001775597,0.8282077,0.002233356,0.01399076,0.00744916,0.1397811],"study_design_scores_gemma":[0.00001201121,0.00003667983,0.0003927053,0.00002791892,0.00001058096,0.00003696477,0.00001215456,0.9889871,0.001136473,0.008899436,0.0004379549,0.000009961435],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1197658,0.001618093,0.8649762,0.002481977,0.0003392534,0.0001816918,0.0007159238,0.00563851,0.004282552],"genre_scores_gemma":[0.8762785,0.000437487,0.1179286,0.0005885622,0.0001275588,0.0001621876,0.001164632,0.0007069728,0.00260542],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.006186683,"threshold_uncertainty_score":0.03234875,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01637108806119735,"score_gpt":0.2768592316950903,"score_spread":0.2604881436338929,"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."}}