SimpleShot: Revisiting Nearest-Neighbor Classification for Few-Shot Learning
This work addresses the problem of overfitting in few-shot learning for researchers, showing that incremental improvements to basic methods can be effective.
The paper tackled few-shot learning by revisiting nearest-neighbor classification without meta-learning, finding that simple feature transformations like mean-subtraction and L2-normalization achieve competitive accuracies, outperforming prior results in three out of five settings on miniImageNet.
Few-shot learners aim to recognize new object classes based on a small number of labeled training examples. To prevent overfitting, state-of-the-art few-shot learners use meta-learning on convolutional-network features and perform classification using a nearest-neighbor classifier. This paper studies the accuracy of nearest-neighbor baselines without meta-learning. Surprisingly, we find simple feature transformations suffice to obtain competitive few-shot learning accuracies. For example, we find that a nearest-neighbor classifier used in combination with mean-subtraction and L2-normalization outperforms prior results in three out of five settings on the miniImageNet dataset.