|
![]() |
![]() |
![]() |
![]() |
![]() ![]() |
![]() |
![]() |
![]() |
Publications | ![]() |
![]() |
Domaines Capteurs et modélisation 3D Réalité virtuelle et simulation Thermographie et vision infrarouge Traitement de sequences video Vision biologique Vision cognitive Projets E-Inclusion COBVIS-D Auto21 SKALPEL-ICT COGNOIS VERTEX VRES Chercheurs Christian Gagné Xavier Maldague Denis Poussart Simon Gagné Marc Parizeau Robert Bergevin André Zaccarin Jean-François Lalonde Abdel Hakim Bendada Dates Trier par date dans l'ordre decroissant |
![]() |
A New Loss Function for Temperature Scaling to have Better Calibrated Deep NetworksAbstract - However Deep neural networks recently have achieved impressive results for different tasks, they suffer from poor uncertainty prediction. Temperature Scaling(TS) is an efficient post-processing method for calibrating DNNs toward to have more accurate uncertainty prediction. TS relies on a single parameter T which softens the logit layer of a DNN and the optimal value of it is found by minimizing on Negative Log Likelihood (NLL) loss function. In this paper, we discuss about weakness of NLL loss function, especially for DNNs with high accuracy and propose a new loss function called Attended-NLL which can improve TS calibration ability significantly. ![]() ![]() Bibtex:
@article{Mozafari1206, Dernière modification: 2018/12/03 par cgagne |
![]() |
![]() |
![]() |
![]() |
|||
©2002-2025. Laboratoire de Vision et Systèmes Numériques. Tous droits réservés |