Consistent Relative Confidence and Label-Free Model Selection for Convolutional Neural Networks

This letter is concerned with image classification with deep convolutional neural networks (CNNs). The focus is on the following question: given a set of candidate CNN models, how to select the right one with the best generalization property for the current task? Present model selection methods require access to a batch of labeled data for computing a pre-specified performance metric, such as the cross-entropy loss, the classification error rate, the negative log-likelihood. In many practical cases, labels are not available in time as labeling itself is a time-consuming and expensive task. To this end, this letter presents an approach to CNN model selection using only unlabeled data. This method is developed based on a principle termed consistent relative confidence. The effectiveness and efficiency of the proposed method are demonstrated by experiments using benchmark datasets.