Skip to yearly menu bar Skip to main content


Poster

Learning to Self-Train for Semi-Supervised Few-Shot Classification

Xinzhe Li · Qianru Sun · Yaoyao Liu · Qin Zhou · Shibao Zheng · Tat-Seng Chua · Bernt Schiele

East Exhibition Hall B + C #21

Keywords: [ Algorithms -> Meta-Learning; Algorithms ] [ Semi-Supervised Learning ] [ Few-Shot Learning ] [ Algorithms ]


Abstract:

Few-shot classification (FSC) is challenging due to the scarcity of labeled training data (e.g. only one labeled data point per class). Meta-learning has shown to achieve promising results by learning to initialize a classification model for FSC. In this paper we propose a novel semi-supervised meta-learning method called learning to self-train (LST) that leverages unlabeled data and specifically meta-learns how to cherry-pick and label such unsupervised data to further improve performance. To this end, we train the LST model through a large number of semi-supervised few-shot tasks. On each task, we train a few-shot model to predict pseudo labels for unlabeled data, and then iterate the self-training steps on labeled and pseudo-labeled data with each step followed by fine-tuning. We additionally learn a soft weighting network (SWN) to optimize the self-training weights of pseudo labels so that better ones can contribute more to gradient descent optimization. We evaluate our LST method on two ImageNet benchmarks for semi-supervised few-shot classification and achieve large improvements over the state-of-the-art.

Live content is unavailable. Log in and register to view live content