Self-Attentive Networks for one-shot Image Recognition
Lists
Only you see this
Abstract
Despite recent breakthroughs in applications of deep neural networks, learning from a limited number of examples remains a key challenge in academia and industry. A prototypical example is one-shot learning, in which we must make predictions when only given one sample of each unseen class. Most previous works either categorize instances by the pair similarities or use Long-Short Term Memory (LSTM) structure based method, however, they usually suffer from inter-and intra-class variation problem or high computation cost. In this paper, we propose self-attentive networks to address the above issues in one-shot image recognition problem. Deep neural features and attention mechanism especially self-attention are employed in the proposed self-attentive networks. Experiments on Omniglot and MiniImagenet datasets show that our proposed self-attentive networks achieve state-of-the-art accuracy on one-shot image recognition. In addition, owing to low time complexity and parallelizable architecture, self-attentive networks require significantly less time to train and infer.
BibTeX
@inproceedings{fang2019self,
author = {Pin Fang and Yisen Wang and Yuan Luo},
title = {{Self-Attentive Networks for one-shot Image Recognition}},
booktitle = {ICME},
year = {2019},
url = {https://openreview.net/forum?id=oazWz28NaH},
}Loading related content...