Few shot learning is an important problem in machine learning as large labelled datasets take considerable time and effort to assemble. Most few-shot learning algorithms suffer from one of two limitations- they either require the design of sophisticated models and loss functions, thus hampering interpretability; or employ statistical techniques but make assumptions that may not hold across different datasets or features. Developing on recent work in extrapolating distributions of small sample classes from the most similar larger classes, we propose a Generalized sampling method that learns to estimate few-shot distributions for classification as weighted random variables of all large classes. We use a form of covariance shrinkage to provide robustness against singular covariances due to overparameterized features or small datasets. We show that our sampled points are close to few-shot classes even in cases when there are no similar large classes in the training set. Our method works with arbitrary off-the-shelf feature extractors and outperforms existing state-of-the-art on miniImagenet, CUB and Stanford Dogs datasets by 3% to 5% on 5way-1shot and 5way-5shot tasks and by 1% in challenging cross domain tasks.
View on arXiv