DISTANCE-BASED LEARNING CONFIDENCE MODEL

A method (500) for jointly training a classification model (210) and a confidence model (220) includes receiving a training data set (110) including a plurality of training data subsets (112) and selecting a support set of training examples (114S) and a query set of training examples (114Q), The method also includes updating parameters of the classification model based on a class distance measure and a ground-truth distance associated with a query encoding (212Q) generated for each training example in the query set of training examples. For each training example identified as being misclassified, the method further includes sampling a new query encoding (224) and updating parameters of the confidence model based on the new query encoding.