• Corpus ID: 235658959

Hyperbolic Busemann Learning with Ideal Prototypes

  title={Hyperbolic Busemann Learning with Ideal Prototypes},
  author={Mina Ghadimi Atigh and Martin Keller-Ressel and Pascal Mettes},
  booktitle={Neural Information Processing Systems},
Hyperbolic space has become a popular choice of manifold for representation learning of arbitrary data, from tree-like structures and text to graphs. Building on the success of deep learning with prototypes in Euclidean and hyperspherical spaces, a few recent works have proposed hyperbolic prototypes for classification. Such approaches enable effective learning in low-dimensional output spaces and can exploit hierarchical relations amongst classes, but require privileged information about class… 

Figures and Tables from this paper

Hyperbolic Sliced-Wasserstein via Geodesic and Horospherical Projections

This work proposes to derive novel hyperbolic sliced-Wasserstein discrepancies based on constructions which use projections on the underlying geodesics either along horospheres or geodesICS, and studies and compares these constructions on different tasks wherehyperbolic representations are relevant.

Few-shot Classification with Hypersphere Modeling of Prototypes

This work uses tensor fields (“areas”) to model classes from the geometrical perspective for few-shot learning and presents a simple and effective method, dubbed as hypersphere prototypes ( HyperProto), where class information is represented by hyperspheres with dynamic sizes with two sets of learnable parameters.

Maximum Class Separation as Inductive Bias in One Matrix

This paper proposes a simple alternative to maximum separation as an inductive bias in the network by adding one matrix multiplication before computing the softmax activations, and illustrates that out-of-distribution and open-set recognition benefit from an embedded maximum separation.

Hyperbolic Image Segmentation

Hyperbolic Image Segmentation opens up new possibilities and practical benefits for segmentation, such as uncertainty estimation and boundary information for free, zero-label generalization, and increased performance in low-dimensional output embeddings.

Hierarchical Explanations for Video Action Recognition

Hierarchical ProtoPNet is an interpretable network that explains its reasoning process by considering the hierarchical relationship between classes by dissecting the input video frames on multiple levels of the class hierarchy.



Lorentzian Distance Learning for Hyperbolic Representations

An approach to learn representations based on the Lorentzian distance in hyperbolic geometry, which makes it appropriate to represent hierarchies where parent nodes minimize the distances to their descendants and have smaller Euclidean norm than their children.

Hyperbolic Neural Networks

This work combines the formalism of Mobius gyrovector spaces with the Riemannian geometry of the Poincare model of hyperbolic spaces to derivehyperbolic versions of important deep learning tools: multinomial logistic regression, feed-forward and recurrent neural networks such as gated recurrent units.

Poincaré Embeddings for Learning Hierarchical Representations

This work introduces a new approach for learning hierarchical representations of symbolic data by embedding them into hyperbolic space -- or more precisely into an n-dimensional Poincare ball -- and introduces an efficient algorithm to learn the embeddings based on Riemannian optimization.

Large-Margin Classification in Hyperbolic Space

Hyperbolic SVM, a hyperbolic formulation of support vector machine classifiers, is introduced and its theoretical connection to the Euclidean counterpart is described, allowing end-to-end analyses based on the inherenthyperbolic geometry of the data without resorting to ill-fitting tools developed for Euclidan space.

Representation Tradeoffs for Hyperbolic Embeddings

A hyperbolic generalization of multidimensional scaling (h-MDS), which offers consistently low distortion even with few dimensions across several datasets, is proposed and a PyTorch-based implementation is designed that can handle incomplete information and is scalable.

Poincaré GloVe: Hyperbolic Word Embeddings

Empirically, based on extensive experiments, it is proved that the embeddings, trained unsupervised, are the first to simultaneously outperform strong and popular baselines on the tasks of similarity, analogy and hypernymy detection.

Differentiating through the Fréchet Mean

The Fréchet mean is applied to the existing Hyperbolic Graph Convolutional Network, replacing its projected aggregation to obtain state-of-the-art results on datasets with high hyperbolicity and ahyperbolic batch normalization method is developed that gives an improvement parallel to the one observed in the Euclidean setting.

Hyperbolic Entailment Cones for Learning Hierarchical Embeddings

This work presents a novel method to embed directed acyclic graphs through hierarchical relations as partial orders defined using a family of nested geodesically convex cones and proves that these entailment cones admit an optimal shape with a closed form expression both in the Euclidean and hyperbolic spaces.

Hyperbolic Visual Embedding Learning for Zero-Shot Recognition

A Hyperbolic Visual Embedding Learning Network for zero-shot recognition that is more robust because the embedding feature in hyperbolic space better represents class hierarchy and thereby avoid misleading resulted from unrelated siblings.

Numerically Accurate Hyperbolic Embeddings Using Tiling-Based Models

This work proposes a new model which uses an integer-based tiling to represent \emph{any} point in hyperbolic space with provably bounded numerical error, and enables us to store the resulting embeddings with fewer bits.