Dr. Lydia Fischer: Rejection and online learning with prototype-based classifiers in adaptive metrical spaces

The rising amount of digital data causes the need for intelligent, automated data processing. Classification models constitute particularly popular techniques from the machine learning domain with applications ranging from fraud detection to advanced image classification tasks. Within this talk, we will focus on so-called prototype-based classifiers, since they offer a simple classification scheme, interpretability of the model in terms of prototypes, and good generalisation performance. We will address a few crucial questions which

arise whenever such classifiers are used in real-life scenarios:

• How can we define a confidence for deterministic prototype-based classifiers? Such a measure is relevant    whenever the costs of an error are higher than the costs to reject an example, e. g., in a safety critical system.

• How can we define proper thresholds for an efficient rejection?
• How can we enhance lifelong learning for a prototype-based classifier with confidences?
• Does a hybrid architecture which combines an offline classifier with an online classifier based on their certainty values perform better than the previous approach?