Label Smoothing in Logistic Calibration of Classifiers
Liina Anette Pärtel
This thesis gives a detailed overview of calibrating probabilistic classifiers with logistic calibration also known as Platt scaling. Experiments were carried out to find better label smoothing parameter values than what is used in logistic calibration. Experiments were carried out on toy datasets, varying the size and distribution of classes. The results also show that the current label smoothing parameter formula for Platt scaling is not the optimal value for any of the chosen datasets. It is also noteworthy that the optimal label smoothing parameter depends on both class size and error rate.
Graduation Thesis language
Graduation Thesis type
Bachelor - Computer Science