
Unser Ansatz ANEDL verwendet evidenzbasiertes tiefes Lernen (EDL) als Outlier-Detektor, um verschiedene Arten von Unsicherheit zu quantifizieren, und führt eine adaptive negative Optimierung ein, um die EDL-Erkennung auf den Datensatz mit Inliers und Outliers abzustimmen.


coremsg

adaptive-negative-evidential-deep-learning-für-open-set-semi-supervised-learning


Adaptive Negative Evidential Deep Learning für Open-Set Semi-Supervised Learning


title_rewrite


SCOMatch addresses the overtrusting issue in prior open-set semi-supervised learning methods by treating out-of-distribution (OOD) samples as a new class and forming a new (K+1)-class semi-supervised learning process.


scomatch-improving-open-set-semi-supervised-learning-by-treating-out-of-distribution-samples-as-a-new-class


SCOMatch: Improving Open-set Semi-supervised Learning by Treating Out-of-Distribution Samples as a New Class
