A Relational Model for One-Shot Classification
This addresses the problem of sample-efficient learning for researchers and practitioners in AI, offering a novel approach to one-shot classification.
The paper tackled one-shot image classification by introducing a deep learning model with relational inductive bias, achieving perfect performance on the Omniglot challenge and exceeding human-level accuracy without data augmentation.
We show that a deep learning model with built-in relational inductive bias can bring benefits to sample-efficient learning, without relying on extensive data augmentation. The proposed one-shot classification model performs relational matching of a pair of inputs in the form of local and pairwise attention. Our approach solves perfectly the one-shot image classification Omniglot challenge. Our model exceeds human level accuracy, as well as the previous state of the art, with no data augmentation.