Multi-Class Classification Using a New Sigmoid Loss Function for Minimum Classification Error (MCE)

A new loss function has been introduced for Minimum Classification Error, that approaches optimal Bayes' risk and also gives an improvement in performance over standard MCE systems when evaluated on the Aurora connected digits database.

Saved in:
Bibliographic Details
Published in2010 International Conference on Machine Learning and Applications pp. 84 - 89
Main Authors Ratnagiri, Madhavi V., Rabiner, Lawrence, Juang, Biing-Hwang
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.12.2010
Subjects
Accuracy
Bayes Risk
Estimation
Hidden Markov models
Kernel
Loss measurement
Minimum Classification Error
Noise
Savage Loss
sigmoid loss
Online AccessGet full text
ISBN1424492114
9781424492114
DOI10.1109/ICMLA.2010.20

Cover

More Information
Summary:A new loss function has been introduced for Minimum Classification Error, that approaches optimal Bayes' risk and also gives an improvement in performance over standard MCE systems when evaluated on the Aurora connected digits database.
ISBN:1424492114
9781424492114
DOI:10.1109/ICMLA.2010.20