In recent years, we witnessed a huge success of Convolutional Neural Networks on the task of the image classification. However, these models are notoriously data hungry and require tons of training images to learn the parameters. In contrast, people are far better learner who can learn a new concept very fast with only a few samples. The plausible mysteries making the difference are two fundamental learning mechanisms: learning to learn and learning by analogy. In this paper, we attempt to investigate a new human-like learning method by organically combining these two mechanisms. In particular, we study how to generalize the classification parameters of previously learned concepts to a new concept. we first propose a novel Visual Analogy Network Embedded Regression (VANER) model to jointly learn a low-dimensional embedding space and a linear mapping function from the embedding space to classification parameters for base classes. We then propose an out-of-sample embedding method to learn the embedding of a new class represented by a few samples through its visual analogy with base classes. By inputting the learned embedding into VANER, we can derive the classification parameters for the new class.These classification parameters are purely generalized from base classes (i.e. transferred classification parameters), while the samples in the new class, although only a few, can also be exploited to generate a set of classification parameters (i.e. model classification parameters). Therefore, we further investigate the fusion strategy of the two kinds of parameters so that the prior knowledge and data knowledge can be fully leveraged. We also conduct extensive experiments on ImageNet and the results show that our method can consistently and significantly outperform state-of-the-art baselines.
Submitted 17 Oct 2017 to Computer Vision and Pattern Recognition
Published 18 Oct 2017