Hyperbolic Image Embeddings
By Valentin Khrulkov, Leyla Mirvakhabova, Evgeniya Ustinova, Ivan V. Oseledets, Victor S. Lempitsky. Computer Vision and Pattern Recognition, 2019
411 citations recorded by the source corpus.
Computer vision tasks such as image classification, image retrieval, and few-shot learning are currently dominated by Euclidean and spherical embeddings so that the final decisions about class belongings or the degree of similarity are made using linear hyperplanes, Euclidean distances, or spherical geodesic distances (cosine similarity). In this work, we demonstrate that in many practical scenarios, hyperbolic embeddings provide a better alternative.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition
View Paper: https://www.semanticscholar.org/paper/cc1cc334897af613e9103de9afe23e1882e0c153
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Jure Leskovec
- Rex Ying
- Christopher R'e
- Ines Chami
- William L. Hamilton
- Bharath Hariharan
- Yoshua Bengio
- Davis Wertheimer
In Computer Vision and Pattern Recognition
- Christopher R'e
- Jure Leskovec
- Luming Tang
- Rex Ying
- Davis Wertheimer
- Bharath Hariharan
- Ines Chami
- Chunhua Shen
Citations in the corpus
- Reinforced Attention for Few-Shot Learning and Beyond (2021, neutral): “70 Hyperbolic ProtoNet [20] ResNet-18 59.”
- Hyperbolic Graph Convolutional Neural Networks (2019, positive): “Hyperbolic geometry has been applied to neural networks, to problems of computer vision or natural language processing [8, 14, 18, 38].”
- Fine-Grained Few-Shot Classification with Feature Map Reconstruction Networks (2020, methodology): “Metric-based methods, on the other hand, aim to learn a completely taskindependent embedding that can generalize to novel categories under a chosen distance metric, such as Euclidean distance [20], cosine distance [7], hyperbolic distance [10], or a distance parameterized by a neural network [21].”
- HypRAG: Hyperbolic Dense Retrieval for Retrieval Augmented Generation (2026, unknown): “…enhances performance in various tasks, including molecular generation (Liu et al., 2019), recommendation (Yang et al., 2021; Li et al., 2021), image retrieval (Khrulkov et al., 2020; Wei et al., 2024; Bui et al., 2025), and knowledge graph embedding (Ganea et al., 2018a; Dhingra et al., 2018).”