Abstract:This paper focuses on metric-based few-shot learning, in which the model adjusts its parameters according to the model performance on the meta-tasks encountered during training to obtain a robust model parameter distribution, and the final model performance may lead to overfitting problems due to the small number of samples used to define the categories. Given such problems, a parameter adaptive adjustment mechanism is introduced in the paper to determine the main source of the current model error and to adjust the gradient backpropagation strength in this way to prevent overfitting problems due to sampling errors during the meta-learning training phase. In the paper, a few-shot image classification algorithm based on cross-attention subspace metric is proposed in the framework of existing few-shot image classification algorithms, using a cross-attention mechanism to correlate information between support set sample features and a query set sample features, and then project them into space separately. Different subspaces represent different categories, and the classification accuracy of the model can be improved by measuring the distance between the samples to be measured and the subspaces while increasing the distance between different subspaces to distinguish the samples of different categories. To verify the effectiveness of the method, the proposed method is validated on Mini-ImageNet and tiered-ImageNet, the common datasets for few-shot images, and compared with the classical few-shot image classification algorithm, and the experimental results show that the classification accuracy is improved on both datasets.

Few-Shot Learning Based on Information Geometry

Few-shot Image Classification Based on Gradual Machine Learning

Few-Shot Learning Based on Deep Learning for Image Classification

Gestalt-Guided Image Understanding for Few-Shot Learning

Recent Advances of Few-Shot Learning Methods and Applications

Only Image Cosine Embedding for Few-Shot Learning

LGSim: local task-invariant and global task-specific similarity for few-shot classification

Knowledge-Based Fine-Grained Classification for Few-Shot Learning.

Few-Shot Image Classification: Current Status and Research Trends

Global Information Embedding Network for Few-Shot Learning

Learning with few samples in deep learning for image classification, a mini-review

Adaptive Poincaré Point to Set Distance for Few-Shot Classification.

DeepEMD: Differentiable Earth Mover's Distance for Few-Shot Learning

Deep Metric Learning for Few-Shot Image Classification: A Review of Recent Developments

Joint Distribution Matters: Deep Brownian Distance Covariance for Few-Shot Classification

Discriminant Space Metric Network for Few-Shot Image Classification

Coarse-to-fine Few-Shot Classification with Deep Metric Learning

Multi-Scale Metric Learning for Few-Shot Learning

Adaptive Attribute Distribution Similarity for Few-Shot Learning

Graph Embedding Relation Network for Few-Shot Learning

Image-to-Class Metric based on Category Traversal for Few-shot Learning