Thinner Latent Spaces: Detecting Dimension and Imposing Invariance with Conformal Autoencoders
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909683241975808 |
|---|---|
| author | Kevrekidis, George A. Ahmad, Zan Maggioni, Mauro Villar, Soledad Kevrekidis, Yannis G. |
| author_facet | Kevrekidis, George A. Ahmad, Zan Maggioni, Mauro Villar, Soledad Kevrekidis, Yannis G. |
| contents | Conformal Autoencoders are a neural network architecture that imposes orthogonality conditions between the gradients of latent variables to obtain disentangled representations of data. In this work we show that orthogonality relations within the latent layer of the network can be leveraged to infer the intrinsic dimensionality of nonlinear manifold data sets (locally characterized by the dimension of their tangent space), while simultaneously computing encoding and decoding (embedding) maps. We outline the relevant theory relying on differential geometry, and describe the corresponding gradient-descent optimization algorithm. The method is applied to several data sets and we highlight its applicability, advantages, and shortcomings. In addition, we demonstrate that the same computational technology can be used to build coordinate invariance to local group actions when defined only on a (reduced) submanifold of the embedding space. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_16138 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Thinner Latent Spaces: Detecting Dimension and Imposing Invariance with Conformal Autoencoders Kevrekidis, George A. Ahmad, Zan Maggioni, Mauro Villar, Soledad Kevrekidis, Yannis G. Machine Learning Differential Geometry Conformal Autoencoders are a neural network architecture that imposes orthogonality conditions between the gradients of latent variables to obtain disentangled representations of data. In this work we show that orthogonality relations within the latent layer of the network can be leveraged to infer the intrinsic dimensionality of nonlinear manifold data sets (locally characterized by the dimension of their tangent space), while simultaneously computing encoding and decoding (embedding) maps. We outline the relevant theory relying on differential geometry, and describe the corresponding gradient-descent optimization algorithm. The method is applied to several data sets and we highlight its applicability, advantages, and shortcomings. In addition, we demonstrate that the same computational technology can be used to build coordinate invariance to local group actions when defined only on a (reduced) submanifold of the embedding space. |
| title | Thinner Latent Spaces: Detecting Dimension and Imposing Invariance with Conformal Autoencoders |
| topic | Machine Learning Differential Geometry |
| url | https://arxiv.org/abs/2408.16138 |