Projections of model spaces for latent graph inference
Graph Neural Networks leverage the connectivity structure of graphs as an inductive bias. Latent graph inference focuses on learning an adequate graph structure to diffuse information on. In this work we employ stereographic projections of the hyperbolic and spherical model spaces, as well as produc...
| Главные авторы: | , , |
|---|---|
| Формат: | Conference item |
| Язык: | English |
| Опубликовано: |
OpenReview
2023
|
| _version_ | 1826310758693076992 |
|---|---|
| author | Posner, H Arroyo, A Sáez De Ocáriz Borde, H |
| author_facet | Posner, H Arroyo, A Sáez De Ocáriz Borde, H |
| author_sort | Posner, H |
| collection | OXFORD |
| description | Graph Neural Networks leverage the connectivity structure of graphs as an inductive bias. Latent graph inference focuses on learning an adequate graph structure to diffuse information on. In this work we employ stereographic projections of the hyperbolic and spherical model spaces, as well as products of Riemannian manifolds, for the purpose of latent graph inference. Stereographically projected model spaces achieve comparable performance to their non-projected counterparts, while providing theoretical guarantees that avoid divergence of the spaces when the curvature tends to zero. |
| first_indexed | 2024-03-07T07:56:42Z |
| format | Conference item |
| id | oxford-uuid:e3338a33-8fa2-437a-8b85-a664241a918b |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T07:56:42Z |
| publishDate | 2023 |
| publisher | OpenReview |
| record_format | dspace |
| spelling | oxford-uuid:e3338a33-8fa2-437a-8b85-a664241a918b2023-08-23T08:31:04ZProjections of model spaces for latent graph inferenceConference itemhttp://purl.org/coar/resource_type/c_5794uuid:e3338a33-8fa2-437a-8b85-a664241a918bEnglishSymplectic ElementsOpenReview2023Posner, HArroyo, ASáez De Ocáriz Borde, HGraph Neural Networks leverage the connectivity structure of graphs as an inductive bias. Latent graph inference focuses on learning an adequate graph structure to diffuse information on. In this work we employ stereographic projections of the hyperbolic and spherical model spaces, as well as products of Riemannian manifolds, for the purpose of latent graph inference. Stereographically projected model spaces achieve comparable performance to their non-projected counterparts, while providing theoretical guarantees that avoid divergence of the spaces when the curvature tends to zero. |
| spellingShingle | Posner, H Arroyo, A Sáez De Ocáriz Borde, H Projections of model spaces for latent graph inference |
| title | Projections of model spaces for latent graph inference |
| title_full | Projections of model spaces for latent graph inference |
| title_fullStr | Projections of model spaces for latent graph inference |
| title_full_unstemmed | Projections of model spaces for latent graph inference |
| title_short | Projections of model spaces for latent graph inference |
| title_sort | projections of model spaces for latent graph inference |
| work_keys_str_mv | AT posnerh projectionsofmodelspacesforlatentgraphinference AT arroyoa projectionsofmodelspacesforlatentgraphinference AT saezdeocarizbordeh projectionsofmodelspacesforlatentgraphinference |