Meta-Relation Networks for Few Shot Learning

We propose a meta-relation network to solve the few shot learning problem, where the classifier must learn to recognize new classes given only few examples from each. Meta-relation networks is based on relation networks and Model-Agnostic Meta-Learning (MAML) training methods, which can be trained end-to-end. After training with MAML algorithm, the meta-relation networks can adapt to learning quickly for a small number of samples from a new task with only a small amount of gradient step. It can also classify new classes of images by calculating the scores between query images and few examples of each new class. In experiments, we will demonstrate that the proposed approach leads to good performance on two few-shot image classification benchmarks.

[1]  Oliver Schulte,et al.  Image Caption Generation with Hierarchical Contextual Visual Spatial Attention , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[2]  Tao Xiang,et al.  Learning to Compare: Relation Network for Few-Shot Learning , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[3]  Joan Bruna,et al.  Few-Shot Learning with Graph Neural Networks , 2017, ICLR.

[4]  Sergey Levine,et al.  Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks , 2017, ICML.

[5]  Gregory R. Koch,et al.  Siamese Neural Networks for One-Shot Image Recognition , 2015 .

[6]  Michael S. Bernstein,et al.  ImageNet Large Scale Visual Recognition Challenge , 2014, International Journal of Computer Vision.

[7]  Li Zhang,et al.  Learning to Learn: Meta-Critic Networks for Sample Efficient Learning , 2017, ArXiv.

[8]  Xu Wei,et al.  Learning Like a Child: Fast Novel Visual Concept Learning from Sentence Descriptions of Images , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

[9]  Richard S. Zemel,et al.  Prototypical Networks for Few-shot Learning , 2017, NIPS.

[10]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[11]  Sanja Fidler,et al.  Predicting Deep Zero-Shot Convolutional Neural Networks Using Textual Descriptions , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

[12]  Hang Li,et al.  Meta-SGD: Learning to Learn Quickly for Few Shot Learning , 2017, ArXiv.

[13]  Tara N. Sainath,et al.  Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups , 2012, IEEE Signal Processing Magazine.

[14]  Hugo Larochelle,et al.  Optimization as a Model for Few-Shot Learning , 2016, ICLR.

[15]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[16]  Oriol Vinyals,et al.  Matching Networks for One Shot Learning , 2016, NIPS.

[17]  Yoshua Bengio,et al.  Show, Attend and Tell: Neural Image Caption Generation with Visual Attention , 2015, ICML.

[18]  Joshua B. Tenenbaum,et al.  One shot learning of simple visual concepts , 2011, CogSci.

[19]  Hong Yu,et al.  Meta Networks , 2017, ICML.

[20]  Richard J. Mammone,et al.  Meta-neural networks that learn by learning , 1992, [Proceedings 1992] IJCNN International Joint Conference on Neural Networks.

[21]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[22]  Daan Wierstra,et al.  Meta-Learning with Memory-Augmented Neural Networks , 2016, ICML.