FP_AINet: Fusion Prototype with Adaptive Induction Network for Few-Shot Learning

Mengping Dong, xue Liu, Chaojun Cen, YANG MI, Zhenbo Li

Abstract: Conventional prototypical network treats all samples equally and does not consider the effects of noisy samples, which leads to a biased class representation. In this paper, we propose a novel Fusion Prototype with Adaptive Induction Network (FP_AINet) for few-shot learning that can learn representative prototypes from a few support samples. Specifically, to address the problem of noisy samples in the support set, an adaptive induction network is developed, which can learn different class representations for diverse queries and assign adaptive scores for support samples according to their relative significance. Moreover, the proposed model can generate a more accurate prototype than comparison methods by considering the query-related samples. With an increasing of samples, the prototypical network is more expressive since the Adaptive Induction Network ignores the relative local features. As a result, a Gaussian-based fusion algorithm is designed to learn more representative prototypes. Extensive experiments are conducted on three datasets: miniImageNet, tieredImageNet, and CIFAR_FS. The experimental results compared with the state-of-the-art few-shot learning methods demonstrate the superiority of FP_AINet.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning