Hyperbolic Vision Transformers: Combining Improvements in Metric Learning
This work addresses metric learning for computer vision tasks, offering a novel approach that improves performance but is incremental in combining hyperbolic geometry with existing transformer architectures.
The paper tackles metric learning by proposing a hyperbolic vision transformer model that maps embeddings to hyperbolic space and optimizes them with a modified pairwise cross-entropy loss, achieving new state-of-the-art performance on four datasets.
Metric learning aims to learn a highly discriminative model encouraging the embeddings of similar classes to be close in the chosen metrics and pushed apart for dissimilar ones. The common recipe is to use an encoder to extract embeddings and a distance-based loss function to match the representations -- usually, the Euclidean distance is utilized. An emerging interest in learning hyperbolic data embeddings suggests that hyperbolic geometry can be beneficial for natural data. Following this line of work, we propose a new hyperbolic-based model for metric learning. At the core of our method is a vision transformer with output embeddings mapped to hyperbolic space. These embeddings are directly optimized using modified pairwise cross-entropy loss. We evaluate the proposed model with six different formulations on four datasets achieving the new state-of-the-art performance. The source code is available at https://github.com/htdt/hyp_metric.