skip to main content


Title: Nested Hyperbolic Spaces for Dimensionality Reduction and Hyperbolic NN Design
Award ID(s):
1724174
NSF-PAR ID:
10343271
Author(s) / Creator(s):
Date Published:
Journal Name:
IEEE Conference on Computer Vision and Pattern Recognition
ISSN:
2163-6648
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Hyperbolic neural networks have been popular in the re- cent past due to their ability to represent hierarchical data sets effectively and efficiently. The challenge in develop- ing these networks lies in the nonlinearity of the embed- ding space namely, the Hyperbolic space. Hyperbolic space is a homogeneous Riemannian manifold of the Lorentz group which is a semi-Riemannian manifold, i.e. a mani- fold equipped with an indefinite metric. Most existing meth- ods (with some exceptions) use local linearization to de- fine a variety of operations paralleling those used in tra- ditional deep neural networks in Euclidean spaces. In this paper, we present a novel fully hyperbolic neural network which uses the concept of projections (embeddings) fol- lowed by an intrinsic aggregation and a nonlinearity all within the hyperbolic space. The novelty here lies in the projection which is designed to project data on to a lower- dimensional embedded hyperbolic space and hence leads to a nested hyperbolic space representation independently useful for dimensionality reduction. The main theoretical contribution is that the proposed embedding is proved to be isometric and equivariant under the Lorentz transforma- tions, which are the natural isometric transformations in hyperbolic spaces. This projection is computationally effi- cient since it can be expressed by simple linear operations, and, due to the aforementioned equivariance property, it al- lows for weight sharing. The nested hyperbolic space rep- resentation is the core component of our network and there- fore, we first compare this representation – independent of the network – with other dimensionality reduction methods such as tangent PCA, principal geodesic analysis (PGA) and HoroPCA. Based on this equivariant embedding, we develop a novel fully hyperbolic graph convolutional neural network architecture to learn the parameters of the projec- tion. Finally, we present experiments demonstrating com- parative performance of our network on several publicly available data sets. 
    more » « less
  2. Abstract

    Hyperbolic metamaterial (HMM) is a unique type of anisotropic material that can exhibit metal and dielectric properties at the same time. This unique characteristic results in it having unbounded isofrequency surface contours, leading to exotic phenomena such as spontaneous emission enhancement and applications such as super-resolution imaging. However, at optical frequencies, HMM must be artificially engineered and always requires a metal constituent, whose intrinsic loss significantly limits the experimentally accessible wave vector values, thus negatively impacting the performance of these applications. The need to reduce loss in HMM stimulated the development of the second-generation HMM, termed active HMM, where gain materials are utilized to compensate for metal’s intrinsic loss. With the advent of topological photonics that allows robust light transportation immune to disorders and defects, research on HMM also entered the topological regime. Tremendous efforts have been dedicated to exploring the topological transition from elliptical to hyperbolic dispersion and topologically protected edge states in HMM, which also prompted the invention of lossless HMM formed by all-dielectric material. Furthermore, emerging twistronics can also provide a route to manipulate topological transitions in HMMs. In this review, we survey recent progress in topological effects in HMMs and provide prospects on possible future research directions.

     
    more » « less
  3. null (Ed.)