Predicting the Accuracy of a Few-Shot Classifier
This addresses a practical challenge for researchers and practitioners in few-shot learning by providing alternatives to validation, though it is incremental as it builds on existing transfer-based methods.
The paper tackles the problem of predicting generalization accuracy in few-shot learning when validation sets are unavailable, proposing measures for supervised, semi-supervised, and unsupervised settings that correlate with classifier performance and can predict generalization with confidence.
In the context of few-shot learning, one cannot measure the generalization ability of a trained classifier using validation sets, due to the small number of labeled samples. In this paper, we are interested in finding alternatives to answer the question: is my classifier generalizing well to previously unseen data? We first analyze the reasons for the variability of generalization performances. We then investigate the case of using transfer-based solutions, and consider three settings: i) supervised where we only have access to a few labeled samples, ii) semi-supervised where we have access to both a few labeled samples and a set of unlabeled samples and iii) unsupervised where we only have access to unlabeled samples. For each setting, we propose reasonable measures that we empirically demonstrate to be correlated with the generalization ability of considered classifiers. We also show that these simple measures can be used to predict generalization up to a certain confidence. We conduct our experiments on standard few-shot vision datasets.