Meta-Learning by Hallucinating Useful Examples

Yu-Xiong Wang, Yuki Uchiyama, Martial Hebert, Karteek Alahari

Abstract: Learning to hallucinate additional examples has recently been shown as a promising direction to address few-shot learning tasks, which aim to learn novel concepts from very few examples. The hallucination process, however, is still far from generating effective samples for learning. In this work, we investigate two important requirements for the hallucinator --- (i) precision: the generated examples should lead to good classifier performance, and (ii) collaboration: both the hallucinator and the classification component need to be trained jointly. By integrating these requirements as novel loss functions into a general meta-learning with hallucination framework, our model-agnostic PrecisE Collaborative hAlluciNator (PECAN) facilitates data hallucination to improve the performance of new classification tasks. Extensive experiments demonstrate state-of-the-art performance on competitive miniImageNet and ImageNet based few-shot benchmarks in various scenarios.

Keywords: few-shot learning, meta-learning