Few Shot Learning methods

Machine learning methods have been very successful in data-intensive applications, but is often constrained when the dataset is small. Recently, Few Shot Learning (FSL) has been proposed to solve this problem. Using prior knowledge, FSL can quickly generalise to new tasks containing only a few samples with supervised information.

In this repository, there is an implementation of two research papers where the authors propose FSL methods for image classification (The 2nd paper proposes an enhancement):

• Prototypical Networks for Few-shot Learning
• Few shot learning with Big-Prototypes